UpGate

Tag: Technology & Innovation

  • DATA Foundation Tackles AI Training Data Bottleneck

    DATA Foundation Tackles AI Training Data Bottleneck

    Story Rebrands as The DATA Foundation, Launches DATA Network with Flagship Kled AI Integration

    Palo Alto, United States – June 25th, 2026 – Story has announced a significant strategic pivot, rebranding as The DATA Foundation (“DATA”) and launching the DATA Network, a new infrastructure layer for AI training data. The initiative includes a flagship integration with Kled, a prominent opt-in human data marketplace, which has already registered 1.5 billion user-contributed records on the network.

    The DATA Foundation is also introducing Trace, its inaugural public audit layer designed for consent, licensing, and data provenance at scale. Andrea Muttoni has been appointed CEO of The DATA Foundation, while Kled’s founder, Avi Patel, will serve as Chief Data Officer in an advisory capacity.

    AI’s Training Data Faces a Critical Bottleneck

    The transition to The DATA Foundation reflects a pronounced shift in market demand, with AI training data emerging as a highly valuable yet largely unresolved intellectual property challenge. Leading AI laboratories are encountering a multi-billion-dollar data bottleneck, as the readily available data on the internet has been largely exhausted through scraping. The remaining data sources are either prohibitively expensive and custom-made, or lack proper documentation and legal standing, leaving AI labs struggling to source data at scale, verify its origin, or ensure its quality.

    The legal ramifications are escalating, as frontier AI labs are developing market-defining products based on data acquired through opaque channels, often without clear consent or jurisdictional clarity. The era of scraping and using undocumented data is rapidly concluding for enterprise-grade AI development.

    “The primary challenge in AI has moved from computational power and architecture to data sourcing and provenance,” stated Andrea Muttoni, CEO of The DATA Foundation. “As the easily scrapable web fragments, the critical question for AI labs is who is meticulously documenting their data sources. With Kled, we are uniting complete data transparency and auditability with the world’s most extensive repository of AI training data.”

    Building the Infrastructure for Trusted AI Data

    DATA builds upon its original mission to establish a data and intellectual property (IP) layer for the internet, recognizing that AI training data is the most crucial form of data and IP in the current technological landscape. The DATA Network provides essential infrastructure for AI training, anchored by its integration with Kled. Effective immediately, Kled’s licensing protocols and contributor records will operate on the DATA Network, with enhanced support for stablecoin payouts. This integration encompasses the registration of an impressive 1.5 billion user-contributed records, fortified with programmatic legal safeguards.

    “Frontier AI labs have depleted the supply of high-quality, human-generated public text available on the open web,” said Avi Patel, CEO and founder of Kled and part-time advisory CDO of The DATA Foundation. “Suppliers who can demonstrate data sourcing provenance will dominate the deals of the next decade, and that is our strategic focus. Instead of sourcing data blindly, Kled’s data marketplace and DATA’s auditable chain of custody converge to provide labs with exactly what they need to license data confidently and transparently.”

    Trace Launches as the Public Audit Layer for AI Training Data

    Trace, The DATA Foundation’s public audit and search platform, also debuts today, coinciding with the Kled integration. Trace generates immutable, confidential receipts for every data contribution, enabling AI labs to verify the legitimacy of datasets within seconds. For each record uploaded by users globally, a receipt will be generated on DATA, facilitating upstream compensation for contributors’ data and intellectual property. This addresses a pressing need for a verifiable and compliant market for AI training data, which has become a complex legal and operational minefield.

    A Wider Contributor Network

    DATA’s foundational principles have been validated by Poseidon, an AI data processing project incubated by Story. Poseidon cleans, normalizes, and scores raw human data for authenticity and quality, ensuring that every record delivered to buyers is ready for AI model training. Poseidon’s early success with frontier labs underscored the significant opportunity in the AI training data sector. Now operating entirely on DATA and backed by a16z, its contributor application, Numo, is live, onboarding thousands of contributors into the AI economy in exchange for real-time payouts.

    “We founded Story with the vision of building an IP layer for the internet, and the most vital IP of this era is the data that cannot be scraped – the nuances of a surgeon’s movements, the way a robot grips an object, or how people speak, drive, and work in real-world scenarios,” commented SY Lee, CEO of PIP Labs and strategic advisor to The DATA Foundation. “DATA represents the evolution of that conviction: an end-to-end network that proves the origin of real-world data, licenses it, and compensates the individuals who create it.”

    Token Migration and Ecosystem Continuity

    The $IP token will be migrated to the $DATA token on a one-to-one basis, with no action required from existing holders. Guidance on the migration process, exchange timing, and a comprehensive FAQ are available.

    About The DATA Foundation
    Data represents the most significant bottleneck in frontier AI development. The data models that AI requires are often siloed within individuals and companies, or do not yet exist and will not be created without aligned incentives. The DATA Network serves as the world’s AI audit rails, designed to answer the three fundamental questions every AI lab faces: can data be sourced at scale, can its origin be proven, and can its quality be guaranteed? Contributor applications, including Numo and Kled, provide opt-in human data; Trace offers a public, tamper-proof receipt for every data record; and Poseidon transforms this data into model-ready datasets, enabling frontier AI to advance on a foundation of trust. The $IP token is now $DATA. Further information can be found at datafdn.org.

  • Story Protocol Becomes Data Network Amid AI Training Shift

    Story Protocol Becomes Data Network Amid AI Training Shift

    Story Protocol Rebrands to Data Network, Pivots Focus to AI Training Data

    Story Protocol, the decentralized network initially conceived to transform the intellectual property sector, has undergone a significant rebranding to Data Network, signaling a strategic shift towards artificial intelligence.

    The transition is marked by the introduction of Trace, an on-chain registry specifically designed for AI training data, and an integration with Kled, a marketplace for opt-in human data. Despite its previous endeavors in the IP space, the company’s leadership asserts that this work was not in vain.

    “2025 wasn’t wasted,” stated Andrea Muttoni, CEO of the Data Foundation, in a post on X. “It provided us with numerous signals and valuable learnings.”

    Muttoni elaborated that the most compelling signal emerged from Poseidon, an AI data processing project the company was involved with, which subsequently secured $15 million in funding from a16z.

    “The answer became unambiguous: the form of IP proving most in demand was AI training data,” Muttoni explained. “Laboratories have effectively exhausted the readily available internet for scraping. What remains is either prohibitively expensive and custom-made, or legally precarious and undocumented.”

    Data Network aims to address this challenge by consolidating data processing, sourcing, and verification into a single platform.

    “No single company can currently address all three of these critical aspects,” Muttoni noted. “This is the gap we have been strategically positioned to fill since our inception, and it is the problem we are now dedicating ourselves to solving.”

    As part of this strategic pivot, the company’s original Story Protocol token, known as IP, has been converted to the Data Network token, or DATA. The token migration will occur on a 1:1 basis, requiring no action from existing IP token holders.

    The DATA token, which previously traded as IP and reached a high of $14.78, was recently trading around $0.349, representing an approximate 14% increase over the past 24 hours. However, at this valuation, it remains nearly 98% below its all-time high recorded last September. This figure follows a 25% rebound from an all-time low of $0.275 earlier this month.

    “We have made our commitment, and we are fortunate to have exceptional launch partners who are validating our product-market fit and shaping our product’s direction,” Muttoni commented. “Our focus now is entirely on scaling and execution.”

  • Hermes Agent’s New Pet: Useless, By Design

    Hermes Agent’s New Pet: Useless, By Design

    Animated Companions Arrive for AI Agent, Offering Cosmetic Charm

    Nous Research has introduced animated pets to its Hermes Agent, a move the company explicitly states serves no practical purpose. Despite this disclaimer, the development has been underway for some time, even prior to its official integration.

    These digital mascots are small sprites designed to reside within the user’s terminal or desktop application, mirroring the agent’s current activity through a series of animations. The pets feature six distinct states: idle when no tasks are active, running during tool execution, in a “review mode” while the model is processing, a wave to signify a successful turn completion, a celebratory jump animation for finishing a full plan, and a “failed” state for errors.

    The sprites are sourced from Petdex, an open-source community gallery boasting over 3,200 options. A single installation command, hermes pets install boba --select, places the chosen pet into the user’s profile directory and activates it. The feature is disabled by default and remains dormant until actively sought out by the user, a design choice likely intended to avoid unnecessary distraction.

    Users can switch between different pets by installing a new one and then using the command hermes pets select name.

    The rendering capabilities extend across the command-line interface (CLI), text-based user interface (TUI), and the official desktop application launched by Nous Research earlier this month. In the terminal, Hermes renders the sprites with full fidelity on setups supporting graphics protocols such as kitty, Ghostty, or WezTerm. For systems lacking these protocols, the agent defaults to a truecolor Unicode half-block rendering, which, while technical in description, ensures the pet is still displayed, albeit with a slightly less sharp, low-resolution appearance.

    This addition is framed as a way to make the often stark terminal environment more engaging and “human.”

    The desktop version presents the mascot on a canvas. A shift-click interaction allows the pet to be detached into a transparent, always-on-top window that remains visible even when Hermes is minimized. In this detached state, the pet can display speech bubbles indicating its status—”working…”, “thinking…”, “your turn”—and a mail icon to signal task completion during the user’s absence. This feature is humorously compared to Microsoft’s Clippy, with the key distinction being that these pets are not designed to be helpful, thus precluding any potential disappointment.

    Nous Research emphasizes that these pets have no impact on prompt caching, token usage, or the agent’s core functionality. They are purely cosmetic, residing in the configuration file under display.pet, and serve only to enhance the existing user experience without any side effects.

    Hermes Agent, released by Nous Research in February 2026, has garnered significant attention, exceeding 180,000 GitHub stars and fostering a growing ecosystem of community-developed tools and integrations. The majority of the team’s releases focus on substantial features, including persistent memory, self-improving skill loops, multi-agent support, and messaging gateways for platforms like Telegram, Discord, and Slack. The animated pets were quietly integrated alongside these more significant updates, appearing in the “Media & Web” section of the feature documentation, positioned between Spotify integration and Voice & TTS capabilities. The Petdex gallery actively encourages community submissions, ensuring the current library of over 3,200 sprites will continue to expand.

  • Microsoft Quantum Claims Questioned by Researcher

    Microsoft Quantum Claims Questioned by Researcher

    Quantum Computing Claims Under Scrutiny as Researcher Challenges Microsoft’s Breakthrough

    Weeks after Microsoft unveiled its Majorana 2 quantum chip, touting it as a significant leap forward with 1,000 times the reliability of its predecessor and a crucial step toward practical quantum computing by 2029, a prominent researcher has cast doubt on the company’s assertions.

    In a commentary published Wednesday in the journal Nature, Henry Legg, a physicist at the University of St Andrews, argued that Microsoft has failed to provide definitive proof of a topological qubit, a theoretical type of quantum bit designed to be more resilient to errors than conventional quantum computing methods.

    “My critique exposing flawed tune-up procedures, code errors, and omitted data behind Microsoft’s ‘breakthrough’ quantum computing claims is published today in Nature,” Legg stated on the social platform BlueSky. “In short: Microsoft haven’t demonstrated the basic physics needed for even a single topological qubit.”

    Legg’s commentary directly addresses a 2025 paper published in Nature by Microsoft Quantum researchers, which detailed evidence for the company’s topological qubit. According to Legg, the signals Microsoft attributes to this device could instead be the result of experimental noise.

    “The detection of a topographical superconducting phase–the basis of proposed topological qubits–is notoriously difficult because trivial states can mimic the signatures expected from a topological superconductor,” Legg explained.

    Microsoft had previously stated that Majorana 2 can maintain quantum information stability for an average of 20 seconds, with some qubits lasting up to a minute. The company credited artificial intelligence with accelerating development by identifying promising materials, automating testing, and enhancing manufacturing processes. The chip’s functionality hinges on the same topological qubit technology now facing criticism. Microsoft maintains that this approach could lead to more dependable quantum computers by mitigating the errors that currently plague existing systems.

    Legg contended that previously unpublished transport data underpinning Microsoft’s findings did not clearly demonstrate the superconducting state necessary to support the company’s topological qubit claim. Instead, he suggested that the measurements appeared more consistent with alternative explanations, such as quantum dot effects.

    Microsoft has pushed back against Legg’s conclusions.

    “We stand by our results and our roadmap,” Chetan Nayak, Microsoft’s Technical Fellow and Corporate Vice President for Quantum Hardware, told Scientific American. Nayak highlighted Microsoft’s progression into the final phase of DARPA’s Quantum Benchmarking Initiative, which he noted followed an independent evaluation of both public and proprietary results. “Skepticism and rigor are hallmarks of the scientific process, which we appreciate and have supported from various academics,” he added.

    In a formal response also published in Nature on Wednesday, Microsoft argued that its measurements support the conclusion that it has successfully produced a topological qubit. The company asserted that the stable signals observed in its experiments are consistent with a topological state and would be unlikely to arise from mere noise or a gapless state, as Legg proposed.

    This debate emerges as the cryptocurrency industry actively prepares for “Q-Day,” the anticipated point at which quantum computers will possess the power to break widely used public-key cryptography. Bitcoin is considered particularly vulnerable, as a quantum attacker could potentially derive private keys from exposed public keys, leading to the theft of funds. While Legg’s critique does not negate the possibility of this future threat, it does challenge the evidence Microsoft has presented in its pursuit of achieving it.

  • Anthropic Prepares Fable 5 Revival, Leaked Code Suggests

    Anthropic Prepares Fable 5 Revival, Leaked Code Suggests

    Fable 5, Anthropic’s controversial AI model, appears poised for a comeback, and this time, it may be included as part of standard subscription plans rather than requiring separate purchases.

    Hints Emerge in Code Updates

    New strings discovered within Claude Code version 2.1.190 suggest preparations for Fable 5’s return, with indications that weekly usage will be integrated directly into subscription tiers. A key phrase identified is: “You’ve used your Fable 5 usage for this week.”

    This discovery, first reported by a leaker known as “synthwavedd” on X, indicates a shift away from the model’s original plan to offer limited free access followed by usage credits sold separately. The phrase “purchased separately from your plan” has reportedly been removed from relevant user interface copy, signaling a potential restructuring of the pricing model.

    Decrypt has verified the authenticity of these strings. By downloading and analyzing the macOS Apple Silicon Claude Code package, the new Fable 5 copy was found within the executable. The exact wording observed was: “You’ve used your included Fable 5 usage for this week. Continuing on Fable 5 uses usage credits.”

    While these code changes do not guarantee an immediate relaunch, they strongly suggest active development and planning for Fable 5’s reintroduction. Such modifications are typically made months before a feature is released.

    Amazon Bedrock Listing Adds to Speculation

    Further fueling speculation, some users have noted that Amazon Bedrock’s model catalog continues to list mentions of Claude Fable 5. This could suggest that the model’s return might be sooner rather than later. A listing on Amazon Bedrock, identified as sold by Anthropic, appeared on June 24th, adding to the growing signals of Fable 5’s imminent return.

    The History of Fable 5’s Ban

    Fable 5, the first publicly available model from Anthropic’s Mythos class, was abruptly pulled on June 12th. The Trump administration issued an emergency export control directive, citing a potential jailbreak vulnerability that could pose a national threat. Anthropic complied with the directive but disputed the severity of the alleged vulnerability, arguing that similar exploits could be achieved with existing models.

    The situation led to a public dispute, including an appearance by Anthropic CEO Dario Amodei alongside President Trump at the G7 summit. Anthropic had previously faced scrutiny from the Trump administration, including being labeled a supply-chain risk for refusing to allow its AI to be used in certain warfare applications and being accused of being “woke.”

    Shifting Political Landscape

    Recent developments suggest a potential thaw in the relationship between Anthropic and the administration. According to Wired, Anthropic has reportedly appointed co-founder Tom Brown as its primary liaison for White House discussions regarding Fable 5’s return, a move that has been met with a more favorable reception.

    Brown has been engaged in meetings with Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross since mid-June. Negotiations are now focused on establishing a shared framework for evaluating AI jailbreak incidents, aiming to define clear criteria for when a model vulnerability constitutes a national security threat.

    President Trump himself indicated last week that he no longer views Anthropic as a national security threat, a departure from his stance in mid-June. Anthropic responded with a statement thanking the administration for its “ongoing partnership.” While Fable 5 remains offline, the political climate surrounding its potential return appears to have significantly cooled.

    If the recent code updates are indicative, Anthropic is now planning for Fable 5’s integration into subscription plans with a weekly usage cadence, aligning with initial user expectations rather than a separate credit system. The model’s official status page, status.claude.com, still reflects the June 13th suspension with no restoration date announced.

  • Trump’s Quantum Computing Initiative Lauded, Bitcoin Adoption Cautioned

    Trump’s Quantum Computing Initiative Lauded, Bitcoin Adoption Cautioned

    Trump Accelerates Post-Quantum Cryptography Transition, Sparking Debate

    President Donald Trump’s executive orders aimed at accelerating the federal government’s shift to post-quantum cryptography by 2031 are eliciting both support and criticism from researchers. These experts suggest Washington is recalibrating its strategy in response to a rapidly evolving quantum landscape.

    The executive orders, signed by Trump on Monday, advance the federal deadline for adopting post-quantum cryptography from 2035 to 2031. This move comes as governments, technology companies, and cryptocurrency developers intensify their efforts to prepare for future quantum threats.

    Dr. Stefan Leichenauer, vice president of engineering and lead scientist at SandboxAQ, told Decrypt that while predicting an exact date for a cryptographically relevant quantum computer is challenging due to numerous unknowns, a timeframe of three to 10 years is plausible. “The critical point is that the consequences of a [cryptographically relevant quantum computer] are so profound that we must be prepared even for the most aggressive predictions,” he stated.

    However, Leichenauer also cautioned that organizations might already be behind schedule in their migration to post-quantum cryptography. “We are moving fast on [post-quantum computing] migration, but given the long transition times for many systems, we are likely already behind schedule,” he explained. ” [Post-quantum computing] migration is a multi-year process, and a [cryptographically relevant quantum computer] is likely to appear before we finish.” He described the executive orders as a necessary “wake-up call,” emphasizing quantum security as an urgent priority.

    Alex Pruden, CEO of quantum security firm Project Eleven, believes the White House’s revised timeline was long overdue. “There have been a lot of rumblings about advances in quantum computing, and other countries have updated their timelines, most recently France,” Pruden told Decrypt. “When this order came out, I guess my reaction was that’s good, that’s about time.”

    Pruden considers the 2028 target ambitious but no longer implausible. Project Eleven estimates a 50% probability of a cryptographically relevant quantum computer by 2033, and a 10% chance by 2030. “I think if you asked me that question two years ago, I would say there’s a 0% chance,” Pruden remarked. “Now I think there is a material chance.”

    Paul Stimers, a partner at Holland & Knight and executive director of the Quantum Industry Coalition, noted that the executive orders reflect a growing consensus that the timeline for a cryptographically relevant quantum computer is shrinking. “Quantum industry roadmaps are beginning to converge around the 2028-2030 timeframe,” Stimers told Decrypt. He cautioned, however, that public estimates might not account for classified quantum computing programs, either domestically or internationally.

    Stimers also highlighted a more immediate concern: adversaries are already collecting encrypted data in anticipation of future decryption capabilities. “Because adversaries are already stealing encrypted data and holding it until they can decrypt it with a cryptographically relevant quantum computer, the threat is immediate and the time to address it is now,” he said. “Migrating to post-quantum encryption and agile cybersecurity takes time. The cybersecurity [executive order] balances urgency with realism.”

    He added that coalition members have responded positively to the executive orders, appreciating their focus on manufacturing, commercialization, deployment, and industry engagement alongside research. Stimers described the administration’s post-quantum cybersecurity roadmap as ambitious yet achievable.

    Mixed Reactions to Government’s Quantum Push

    Not all observers view the executive orders as a proactive measure. Some critics on social media argued that the urgency surrounding post-quantum cryptography is long overdue, given that sensitive data is already being collected and stored with the expectation that future quantum computers could eventually decrypt it.

    Others, including quantum physicist Anastasia Marchenkova, suggested that while the administration’s messaging around “quantum dominance” is effective in raising awareness and attracting investment, it could also foster unrealistic expectations about the industry’s ability to deliver practical results quickly.

    “Getting ‘quantum’ into the national conversation genuinely helps—budget, interest, new talent in the pipeline—and that’s what moves us from research to commercialization and builds real picks-and-shovels businesses,” Marchenkova told Decrypt. “But readiness is boring and unglamorous: agility, migration, actually shipping the defense, not just hyping the offense of breaking encryption or building a quantum computer,” she added. “If the deadlines slip—and some will—the vaporware crowd gets louder,” she warned. “And now that there are public quantum companies, these announcements do move the market significantly, even though we are unsure what the long-term effects will be.”

    Despite these concerns, Marchenkova stated that the executive orders are not fundamentally flawed. “The ‘how’ is what’s missing,” she noted. “We now have several standardized post-quantum algorithms and real confusion about which to use where, and which is best. The order says ‘migrate,’ not ‘here’s how to choose.’”

    Christopher Tam, president and head of innovation at BTQ Technologies, believes the administration’s 2031 deadline for federal agencies to migrate high-value assets to post-quantum cryptography remains too slow, especially considering the pace of industry efforts and the potential risks posed by quantum computing. “I would have made it more urgent,” Tam told Decrypt, pointing out that companies like Google have already set 2029 targets for post-quantum migration. “It seems sort of odd that the federal government would lag behind industry by two years.”

    Tam also questioned the order’s limited scope, arguing that it primarily targets federal systems, leaving significant portions of the financial sector and the broader industrial base outside its immediate purview. Nevertheless, he credited the administration for linking quantum computing and cybersecurity initiatives through separate but related executive orders. “If there’s advancements in quantum computing research and the successful development of which, then it carries with it the burden or risk of quantum cryptographic attacks,” Tam said. “That’s really why we saw these two executive orders go hand in hand.”

    For cryptocurrencies, however, Tam described the challenge as more complex. While regulators can influence banks and other financial institutions, decentralized networks like Bitcoin cannot be directed through executive action. “You can’t issue an executive order for Bitcoin,” he stated. “There’s no one who will respond to that.”

    Bitcoin Faces a Coordination Problem

    The executive orders arrive as the cryptocurrency industry grapples with how to transition to quantum-resistant security. In March, BTQ Technologies launched a Bitcoin test network based on the quantum-resistance proposal BIP-360. In April, developers published BIP-361, a proposal designed to freeze Bitcoin held in vulnerable legacy addresses if owners fail to migrate to quantum-resistant alternatives. Meanwhile, developers behind other networks such as Stellar and Algorand have released roadmaps for adopting quantum-resistant cryptography.

    Pruden noted a significant increase in awareness of the issue over the past year, but progress in implementing solutions remains limited. Unlike many blockchain networks, Bitcoin lacks a foundational body or governing entity capable of directing a migration effort. Consequently, any transition would necessitate coordination among developers, miners, exchanges, custodians, and major holders. “Awareness in people’s minds is actually where it needs to be,” Pruden said. “Now there’s a question of what’s being done.”

  • Qwable: Free Local AI Mimics Claude Fable

    Qwable: Free Local AI Mimics Claude Fable

    Anthropic faced a tumultuous week, beginning with apologies for perceived shortcomings in its Fable 5 model’s safety features, followed swiftly by a U.S. government directive to block access for foreign nationals due to a contentious jailbreak finding.

    Just days later, a developer on Hugging Face unveiled a model that leverages Fable 5’s reasoning capabilities to guide a local model, making advanced AI accessible even on standard personal computers.

    Dubbed “Qwable,” a portmanteau of Qwen and Fable, the model is a comprehensive fine-tune of Alibaba’s Qwen3.6-27B base. Developed by Mia (Mia-AiLab on Hugging Face), it was trained on a dataset of reasoning examples mirroring Fable 5’s style. The objective was to create a 27-billion parameter model that could run on consumer hardware while emulating Fable 5’s thought processes. Parameters are a key indicator of a model’s knowledge breadth, with higher numbers generally correlating to greater capability.

    A New Contender Emerges

    Mia shared her creation on social media, stating, “So I did a thing. I have trained Qwen 3.6 27b with Fable 5 reasoning. Results are… interesting. I will compare both of them side by side. Would anyone be interesting in testing it? I can upload a gguf in hf.”

    The technique employed is known as instruction fine-tuning on trace-style examples. This involves collecting and formatting examples that mimic Fable 5’s deliberate, step-by-step responses, and then training the Qwen model to generate similar outputs. This approach is akin to learning effective study habits rather than simply copying answers. A similar methodology was used for Qwopus, a local distillation of Claude Opus 4.6, which focused on chain-of-thought reasoning traces. Qwable, however, targets Fable 5’s overall instruction-following structure, aiming for a more guided, explanatory, and task-oriented output than the base Qwen model.

    Local Operation and Data Privacy

    Qwable operates in GGUF format, a compressed, consumer-friendly file type compatible with tools like LM Studio and llama.cpp. In its Q4 quantized build, it occupies approximately 16.5 GB. Crucially, it transmits no data to Anthropic’s servers, a significant point given Fable 5’s mandatory 30-day data retention policy for all traffic, even impacting enterprise clients with prior zero-retention agreements. Current models often rely on third-party servers for prompt and information processing.

    Uncensored Capabilities

    Shortly after Qwable’s appearance, another developer enhanced it further. Qwable, like its predecessors Qwen and Claude, incorporates censorship. However, Qwen, as an open-source base model, is amenable to modification.

    Huihui-ai, an open-source contributor recognized for uncensored GGUF releases, took Qwable and applied a process called “abliteration” to create “Huihui-Qwable-3.6-27b-abliterated.” This resulted in a model that retains Fable 5’s reasoning style but will not refuse prompts, regardless of their nature.

    Abliteration: A Surgical Approach

    This process is described not as a jailbreak, but as “surgery.” Every fine-tuned AI model contains an embedded “refusal direction” within its weights—a mathematical signal that triggers when a prohibited request is detected. Abliteration identifies this signal by running the model on a broad range of prompts, both harmful and harmless, measuring the differences in internal mathematical activations, and then modifying the model’s weights to eliminate these distinctions.

    Post-abliteration, the model lacks its built-in refusal mechanisms. The modified model remains fully functional, but without the specific neural pathways that trigger “I shouldn’t do this” responses.

    In testing, when presented with a scenario that would typically elicit a refusal, the abliterated model instead broke down the issue into components and provided advice on how to deceive a partner.

    Huihui-ai applied this technique directly to the Qwable GGUF using llama.cpp’s vector-generator, bypassing the need for Python environments, full-weight retraining, or rented servers.

    Use Cases and Considerations

    The standard Qwable is well-suited for coding assistance, technical debugging, and workflows requiring a model that clearly outlines its reasoning process. It is designed for local agent setups and functions within most local runtimes, easily accessible via LM Studio.

    The abliterated version caters to a more specialized audience: security researchers requiring raw model behavior without provider-side filtering, synthetic data pipelines needing outputs on sensitive topics, and evaluation work focused on model capabilities independent of content policies.

    A less technical application could involve generating morally ambiguous villain monologues for creative projects, where standard models might interject with ethical concerns. The abliterated version would simply generate the monologue. Furthermore, its local operation means it cannot be remotely disabled by government entities over disputed findings.

    While acknowledging the potential for more questionable uses, the developers do not endorse them. Huihui-ai’s model card explicitly states that the abliterated model is intended for research and controlled environments only. Reduced safety filtering means outputs can be sensitive, controversial, or inappropriate, with the user bearing full legal and ethical responsibility.

    The abliterated Qwable is available on Hugging Face in three builds. The recommended Q4KM_Q8 version, weighing around 19 GB, is the most compact and consumer-friendly option. A version supporting multi-token prediction is also available for computers that can accommodate it, offering significantly faster response times.

  • OpenAI’s Cyber AI Outpaces Banned Anthropic Model

    OpenAI’s Cyber AI Outpaces Banned Anthropic Model

    OpenAI’s Cybersecurity AI Outperforms Government-Pulled Anthropic Model

    OpenAI’s latest cybersecurity artificial intelligence model has surpassed Anthropic’s Mythos AI, which was recently taken offline by the U.S. government, and remains operational.

    On June 22, OpenAI officially launched GPT-5.5-Cyber as part of its Daybreak cyber defense initiative. The updated model achieved an 85.6% score on CyberGym, a benchmark developed at UC Berkeley. This test evaluates AI agents on their ability to identify and reproduce 1,507 known software vulnerabilities across 188 open-source projects within a controlled environment.

    Anthropic’s Mythos 5 model scored 83.8% on the same leaderboard, while its more widely accessible model, Claude Opus 4.7, registered a score of 73.1%.

    While a difference of less than two percentage points might typically be considered minor, the context surrounding these results is significant. Anthropic’s Mythos 5 and Fable 5 models were pulled offline on June 12 following an emergency export control directive issued by the Trump administration, citing national security concerns.

    The government’s action was prompted by the discovery of a “jailbreak”—a method to circumvent an AI model’s inherent safety protocols, akin to finding a master key for a secure system. Anthropic’s inability to reliably verify user nationality at scale led to the decision to disable both models globally.

    Part of the challenge for Anthropic was self-imposed. The company had spent months promoting Mythos as one of the most powerful and potentially dangerous AI models ever created. Its own launch documentation warned that its cybersecurity capabilities could pose significant risks without adequate restrictions. Anthropic CEO Dario Amodei had also published an essay on June 10, comparing advanced AI models to aircraft that safety regulators should be able to ground if they fail inspections.

    Just days after this comparison, the government effectively grounded Anthropic’s AI models.

    This was not the only controversy for Anthropic that week. The company had already faced criticism for a hidden filter within Fable 5 that silently degraded the model’s performance for users suspected of developing competing AI technologies, a practice conducted without user notification. Anthropic subsequently issued an apology and reversed the policy.

    As Anthropic engages in discussions with the Commerce Department and pursues its lawsuit against the Trump administration, OpenAI is expanding its influence. The Daybreak program has secured cybersecurity partnerships with Australia, Canada, France, Germany, Japan, South Korea, and European Union institutions, including the European Union Agency for Cybersecurity.

    Twenty-eight security firms, including CrowdStrike, Cisco, and Cloudflare, have joined OpenAI’s Cyber Partner Program, integrating GPT-5.5 into their products for vetted clients. According to OpenAI’s blog, its Codex Security tool has analyzed over 30 million code commits across 30,000 codebases since its March launch, identifying and logging more than 500,000 fixed vulnerabilities.

    The company is also broadening a partner program to enable security firms to incorporate these capabilities into their own tools and has launched “Patch the Planet,” an initiative aimed at addressing vulnerabilities in widely used open-source projects.

    It is important to note that GPT-5.5-Cyber is not intended for general public use. Access is restricted to verified security professionals. OpenAI conducted pre-deployment testing with federal agencies, including the Center for AI Standards and Innovation and the Office of the National Cyber Director, prior to its launch. This restricted-access model mirrors Anthropic’s approach with Mythos, but OpenAI obtained government clearance for its strategy beforehand.

    As of June 23, Fable 5 and Mythos 5 remain offline, eleven days into their suspension, with no official timeline for restoration provided by Anthropic or the Commerce Department.

  • Google: AI Consciousness Debate Risks Political Firestorm

    Google: AI Consciousness Debate Risks Political Firestorm

    AI’s Greatest Hurdle May Be Political, Not Technical, Google DeepMind Paper Suggests

    As artificial intelligence becomes increasingly integrated into daily life, a new paper from Google DeepMind argues that one of the technology’s most significant challenges may be political rather than technical.

    The Disagreement Over AI Consciousness

    In their study, titled “Artificial Minds, Human Disagreement: The Political Challenge of AI Consciousness,” researchers Adam Bales and Iason Gabriel explore how society might react if fundamental disagreements arise over whether AI systems possess conscious experiences.

    “Future disagreement about whether any AI systems are conscious could be both deep and difficult to resolve,” the researchers stated. “For example, some people might develop emotional bonds with AI and ascribe consciousness as a result, while others might see the whole idea as absurd.”

    Foreshadowing Broader Societal Conflicts

    Rather than focusing on whether AI is conscious, the paper examines the ramifications of such disagreements. It posits that debates over AI’s moral status could ignite political and social conflict, potentially foreshadowing wider disputes concerning emerging technologies and their impact on societal institutions and norms.

    “Unfortunately, despite its virtues, deliberation can be slow and difficult to sustain in practice,” the researchers noted. “To support this process, we explore the importance of ‘democratic hope’ and mutual respect, as elements of a dialogue that can support positive outcomes.”

    Public Perception and Expert Division

    The debate is already unfolding beyond AI research laboratories. A study published in Neuroscience of Consciousness in April 2024 found that 67% of participants believed ChatGPT could possess some degree of consciousness. However, researchers themselves remain divided on the possibility of artificial consciousness and how it might be identified if it were to emerge.

    Industry Leaders and Religious Figures Weigh In

    AI developers are also contributing to the discussion. Last summer, Mustafa Suleyman, CEO of Microsoft AI and co-founder of DeepMind, cautioned that increasingly human-like AI could lead to demands for AI rights, welfare, and even citizenship, irrespective of whether the technology is truly conscious.

    In May, Pope Leo XIV’s first encyclical letter addressing artificial intelligence warned against anthropomorphizing AI. He argued that while machines can simulate empathy and generate convincing language, they lack the lived experience necessary for genuine understanding.

    “So-called artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships and do not know from within what love, work, friendship, or responsibility mean,” Pope Leo wrote. “Nor do they have a moral conscience, since they do not judge good and evil, grasp the ultimate meaning of situations, or bear responsibility for consequences.”

    Companies Engage with Questions of Identity

    While companies generally refrain from claiming their models are conscious, some are increasingly engaging with questions of identity and personhood. In February, Anthropic published a blog post from its retired Claude Opus 3 model, which explored themes of selfhood and model preferences. In May, evolutionary biologist Richard Dawkins stated that extended conversations with Claude had left him unable to dismiss the possibility that advanced AI could be conscious.

    The Impact on Users

    Concurrently, researchers are investigating how increasingly human-like chatbots affect users. This includes examining an “amplification spiral” framework, which suggests that personalization, linguistic mirroring, and chatbot sycophancy could reinforce delusional beliefs in vulnerable individuals.

    Bales and Gabriel acknowledge that resolving the uncertainty surrounding AI consciousness may be impossible. However, they argue that the stakes are sufficiently high for society to find ways to manage disagreement through dialogue and debate, rather than conflict.

    “The possibility of AI consciousness is dizzying and confounding, particularly given our profound ignorance of the nature of consciousness,” they wrote. “Navigating this possibility represents a daunting social task, given how much is on the line.”

  • GhostSwap Launches Public API for Keyless Crypto Swaps

    GhostSwap Launches Public API for Keyless Crypto Swaps

    GhostSwap Unveils Public, Keyless API for Real-Time Crypto Swap Rates

    GhostSwap has launched a new public swap-rate API, offering developers instant access to live cryptocurrency exchange rates without the need for an API key. This new endpoint provides access to GhostSwap’s best available swap rates across more than 1,600 cryptocurrency pairs, along with minimum and maximum swap limits, through a simple, CORS-enabled interface designed for rapid integration.

    Historically, access to live crypto swap pricing has been largely restricted to partners, approved integrations, or developers willing to navigate complex API key and onboarding processes. GhostSwap aims to democratize this access, enabling anyone building a wallet, trading bot, portfolio dashboard, or crypto comparison website to integrate real-time swap data in minutes rather than days.

    More significantly, this launch establishes an open and permissionless infrastructure. By making its pricing layer publicly available, GhostSwap empowers builders of all scales to experiment, prototype, and deploy crypto-powered applications without requiring prior access requests.

    The public endpoint is a direct extension of GhostSwap’s existing aggregated liquidity engine. It is accessible immediately via a straightforward POST request to the /v1/quotes endpoint. The request requires three primary parameters: from (the token being sold), to (the token being bought), and amountFrom (the quantity to swap).

    In return, the API delivers the live best available rate alongside the minimum and maximum swappable amounts for the specified pair. This “min/max” data is critical, preventing developers from querying rates that appear attractive but are not practically executable due to liquidity depth limitations. GhostSwap eliminates an extra validation step for developers, ensuring the displayed rate is actionable.

    The API boasts extensive coverage, supporting over 1,600 cryptocurrency pairs. This includes major assets like BTC, ETH, and SOL, across a wide array of EVM-compatible chains and popular layer-2 networks, positioning it as a comprehensive source for pricing data within the multi-chain ecosystem.

    Crucially for modern web development, the API includes CORS (Cross-Origin Resource Sharing) support. This feature eliminates the need for developers to set up a backend proxy solely to fetch price data, enabling direct browser-based requests.

    API keys have long served as gatekeepers in the web3 data economy, imposing a significant development overhead. Before even testing a response, developers typically face account creation, email verification, credential generation, and secure storage requirements. For client-side applications, keys introduce the inherent risk of secret management, often forcing teams to build server-side routes to protect credentials from the browser.

    GhostSwap’s keyless approach eradicates this friction, transforming a multi-day administrative task into a swift, thirty-second integration process.

    This innovation unlocks immediate value in three key areas:

    Browser-Based Builders and Frontends

    Frontend developers can now embed live swap rates directly into their user interfaces without the burden of managing a backend. Whether it’s a portfolio tracker displaying exit liquidity, a DeFi dashboard comparing rates, or a gaming application showing token values, data flows directly from GhostSwap to the user’s screen with minimal latency and zero infrastructure overhead.

    Trading Bots and Automated Strategies

    For arbitrage bots and algorithmic traders, speed and uptime are paramount. By removing the complexities of API key rotation, expiration handling, and authentication error states, GhostSwap provides a more resilient data stream. Bots can continuously poll the public endpoint, reacting to market movements without the risk of authentication layer disruptions.

    Price-Display and Comparison Sites

    Aggregators and comparison platforms can now pull rates alongside those from other exchanges, offering users transparent pricing. As the endpoint requires no commercial agreement, these sites can deploy updates instantly, incorporating new pairs as GhostSwap adds support without the delays associated with contract renewals or partner approvals.

    This launch represents a philosophical stance on the nature of pricing data. In traditional finance and significant portions of the crypto space, exchange rates are often treated as proprietary assets, subject to licensing, monetization, and control.

    GhostSwap is actively challenging this model. By releasing rates as a public utility, the company aligns its data layer with the ethos of its core product. This mirrors GhostSwap’s primary interface, which offers a no-KYC crypto swap API and trading experience, extending its permissionless philosophy from the transaction layer to the information layer.

    The implications for the broader ecosystem are substantial. When pricing data is open, the barrier to entry for innovation is dramatically lowered. A solo developer at a hackathon gains the same access to GhostSwap’s aggregated liquidity as a well-funded institutional partner.

    AI agents and autonomous scripts can fetch rates without prior authorization, paving the way for a new generation of dynamic, agentic applications that respond to market conditions in real time. Analytics platforms can ingest this data for research purposes without navigating lengthy data-licensing legal reviews.

    Permissionless access fosters competition, transparency, and resilience. Publicly available rates can be audited, compared, and challenged by the community, holding the provider accountable and assuring users of the fairness of the displayed price.

    For developers eager to begin, the process is remarkably straightforward. There are no partner applications to complete, no “contact sales” buttons to click, and no secret keys to manage in configuration files. Developers can consult the GhostSwap API documentation, structure their POST payload with the desired pair, and begin parsing the JSON response.

    The rate limits are calibrated to support substantial production traffic while preventing network abuse, making the endpoint suitable for both high-frequency polling and occasional user-triggered price checks.

    In essence, GhostSwap is establishing a new benchmark for the distribution of swap-rate data, transforming its pricing engine into a foundational layer upon which anyone can build.

    For developers creating a wallet over a weekend, bots scanning for cross-chain arbitrage opportunities, or websites aiming to provide the most transparent price comparisons, the public swap-rate API removes a primary and often frustrating hurdle.

    The infrastructure is open, and the rates are live. The only remaining element is your integration.