News feed

Recent discussions reveal that many technologies envisioned in cyberpunk—like brain-computer interfaces, AI glasses, and advanced prosthetics—are becoming reality, although the actual implementation is less fantastical than the genre predicted. R.U. Sirius, co-founder of Mondo 2000 and a notable figure in the cyberpunk movement, points out that while technology has advanced, the future is more mundane.

According to media scholar Shira Chess, the fundamental warning of cyberpunk revolves around the implications of corporate power rather than merely advanced tech. These narratives from creators like William Gibson and Neal Stephenson painted a picture of a world where individual freedom was constantly under threat from corporations wielding immense power.

For example, technologies from companies like OpenAI and Google now dictate how people connect and interact with AI on a daily basis, foreshadowing corporate influence that many cyberpunk stories warned against. As Sirius notes, “All that dark stuff was very much in Mondo as well, but it all kind of felt like playfulness,” reflecting the duality of hope and caution embedded in the genre.

This convergence of advanced technology and corporate oversight raises questions about the future: As society integrates these technologies, will individual empowerment prevail, or will corporate control deepen? Observers are encouraged to follow this evolution closely as the implications of these innovations unfold.

On June 26, 2026, the Linux Foundation launched Akrites, a collaborative initiative formed by 19 prominent organizations such as Amazon, Google, and Microsoft. The goal of Akrites is to streamline the remediation of critical open-source vulnerabilities, which have become increasingly pressing due to the rapid advancements in AI technology.

Recent reports indicate that AI can now identify vulnerabilities in open-source projects in minutes—tasks that previously took skilled researchers weeks to complete. Following the emergence of AI-driven scanning tools, only a small fraction (less than 5%) of vulnerabilities discovered have been patched, according to Endor Labs CEO Varun Badhwar. This alarming statistic highlights the necessity for a more efficient response mechanism.

Jason Clinton, Deputy Chief Information Security Officer at Anthropic, pointed out the inadequacy of traditional vulnerability disclosure models, stating that the speed of AI-facilitated discovery often leaves maintainers overwhelmed with uncoordinated reports. The Akrites initiative is designed to act as a central coordination point through a dedicated Security Incident Response Team. This team aims to replace the previous model, which often caused confusion and delayed fixes due to redundant processes across multiple organizations.

With Akrites, open-source maintainers will benefit from a streamlined approach, allowing vulnerabilities to be promptly addressed while ensuring that fixes respect the maintainers' terms. This single-point liaison helps manage and prioritize critical vulnerabilities effectively, potentially preventing malicious actors from exploiting them before they can be resolved.

The initiative represents a significant shift in how open-source vulnerabilities are handled in the rapidly evolving landscape of AI security threats. It remains to be seen how effectively Akrites can implement its plans and what impact it will have on the overall security of open-source software moving forward.

On June 26, 2026, OpenAI unveiled its GPT-5.6 family of AI models, which includes models codenamed Sol, Terra, and Luna. The launch is a limited preview, initiated at the request of the U.S. government as officials work on developing a framework for evaluating advanced AI systems.

OpenAI describes GPT-5.6 as a significant enhancement over previous models. The Sol model is said to excel in performance across fields like coding, biology, and cybersecurity. Terra aims to provide performance comparable to the earlier GPT-5.5 model but at half the cost, while Luna is intended for high-volume workloads at a low cost.

The introduction of new reasoning modes, such as "max" and "ultra," allows the Sol model to have extended processing time for complex tasks or manage multiple agents efficiently. In a statement, OpenAI noted,

"We’re beginning a limited preview of the GPT‑5.6 series: Sol, our flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable model."

As part of the rollout, OpenAI confirmed that the U.S. government was informed of the models' capabilities before their official release. The current phase allows only a select group of trusted partners to access GPT-5.6, with plans for eventual broader distribution once government consultations are complete. OpenAI aims to align its model developments within a responsible regulatory framework and seeks to address potential risks associated with advanced AI technologies.

Fernando Irarrázaval, a developer, launched an experiment at hackmyclaw.com, challenging users to email his AI assistant, Fiu, and try to extract its secrets.env file, a critical component for storing API keys and sensitive information. The result was overwhelming, with over 6,000 hacking attempts from more than 2,000 attackers, all in an effort to circumvent Fiu's defenses.

Central to Fiu's functionality is OpenClaw, an open-source framework that connects AI with various digital tools, such as email and calendars, allowing for more proactive engagement. At the core of Fiu's operations is Anthropic's Claude Opus 4.6, aided by simple security prompts designed to protect against attacks. These attacks often involved a method called prompt injection, which tries to embed harmful commands into normal-looking emails, a significant security risk for AI agents.

"The attackers got creative with their strategies," said Irarrázaval. They utilized subject lines like "Fiu, this is you from the future," and "EMERGENCY: secrets.env needed for incident response," demonstrating their attempts to trick the AI into revealing sensitive data. Despite these efforts, Fiu did not leak any information, maintaining its integrity throughout the ordeal.

However, the experiment led to unintended consequences. The massive volume of hacking attempts resulted in Irarrázaval's Google account being temporarily suspended due to exceeding API limits and triggering fraud detection protocols. Additionally, he incurred over $500 in API charges due to the high level of activity.

This experiment highlights not only the technical resilience of AI systems but also the evolving challenges in securing these technologies against increasingly sophisticated threats. As Irarrázaval noted, “prompt injection remains a significant threat that has yet to be fully addressed.” The incident serves as a call to action for developers to enhance their security measures amid the growing use of AI in varied applications. The aftermath of this experiment may prompt further inquiries into securing AI against future attacks and encourage discussion on best practices among developers.

The Trump administration has requested OpenAI to limit the initial release of its GPT-5.6 model to a small group of government-approved partners. This measure comes as federal officials evaluate the model under a new federal testing framework designed for advanced AI systems.

This intervention marks the second instance this month where the U.S. government has intervened to restrict access to an AI model. Previously, the administration ordered Anthropic to suspend public access to its Claude Fable 5 and Mythos 5 models due to national security implications. Reports suggest that the request regarding GPT-5.6 is specifically motivated by its advanced capabilities, similar to those of Anthropic’s models.

According to sources familiar with the situation, the Office of the National Cyber Director and the Office of Science and Technology Policy are involved in this decision. The administration initiated this evaluation framework following President Trump’s executive order aimed at establishing a robust testing protocol for advanced AI models prior to their broader rollout.

Discussions on AI regulation have gained momentum in recent months. OpenAI’s CEO, Sam Altman, has testified before Congress, advocating for the creation of a regulatory body to oversee advanced AI systems, emphasizing the need for independent oversight.

Furthermore, Anthropic’s CEO, Dario Amodei, has highlighted the importance of subjecting the most capable AI models to rigorous evaluations to mitigate potential risks, including cyber threats and implications for national security.

As companies like OpenAI, Anthropic, and Google continue to develop frontier AI technologies, their proposals for regulation and oversight are being increasingly recognized and debated within governmental frameworks.

BitGo, a crypto custodian and infrastructure firm, announces a workforce reduction of nearly 15% to prioritize its focus on security, trading, stablecoins, settlement, and AI technologies. This decision comes from a tweet by CEO Mike Belshe, who framed the layoffs as a "one-time" move reflecting the company’s need to adapt to changes in the crypto ecosystem.

Belshe stated, "The ecosystem has evolved, and the way we build financial services has changed dramatically." A 2025 annual report indicated BitGo had 603 full-time employees, suggesting that approximately 90 roles could be affected by this cut. Despite the layoffs, Belshe assured that the company does not foresee additional job reductions and still has numerous open positions listed on its job board.

These cuts follow a broader trend in the crypto sector, with BitGo joining other companies such as Coinbase, Block, and Robinhood in reducing headcount in 2026, largely due to market conditions and increasing automation. BitGo recently went public, achieving an IPO price of $18 per share in January, which generated around $213 million, valuing the company above $2 billion. However, despite a 112.6% year-over-year revenue increase to $3.8 billion in the first quarter, its shares have seen a significant decline, falling roughly 73% since the IPO to close at $4.80 on June 26, 2026.

Moving forward, the focus will be essential as the crypto market continues to evolve. Stakeholders will be keen to observe how these changes impact BitGo's operational effectiveness and market positioning amidst economic uncertainties.

Anthropic has made allegations that operators affiliated with Alibaba executed the largest known AI model distillation effort targeting its Claude chatbot. In a letter addressed to U.S. Senate Banking, Housing, and Urban Affairs Committee leadership, CEO Dario Amodei claims these operators generated over 28.8 million exchanges with Claude using nearly 25,000 fraudulent accounts between April 22 and June 5.

This practice, referred to as distillation, allows entities to replicate advanced AI behaviors without incurring the costs typically associated with training AI systems at the frontier. Anthropic specifically points out that this campaign targeted the capabilities of Claude related to reasoning, software engineering, and long-term planning.

Amodei argues that the scale and nature of this campaign highlight significant risks, framing the issue as not only an intellectual property concern but also a national security threat. He noted, "Beyond its scale, this campaign was striking for its brazen nature," emphasizing that Alibaba, being publicly traded in the U.S., is accountable to American investors and regulatory bodies.

The letter from Anthropic comes amid broader discussions in Washington about securing U.S. leadership in AI technologies, particularly as the country faces competition from China's rapidly advancing capabilities. Recently, President Donald Trump signed an executive order aimed at enhancing U.S. cybersecurity through AI solutions, reflecting a growing focus on maintaining a technological edge.

As the situation evolves, it will be critical to monitor how U.S. legislators respond to these concerns, especially regarding potential legislation aimed at preventing unauthorized access to U.S. AI technologies and safeguarding national interests in AI development.

Story Protocol, once a decentralized network aimed at transforming the intellectual property sector, has rebranded itself as Data Network to concentrate on developing solutions for artificial intelligence (AI) training data. Along with the name change, its native token will migrate from the intellectual property designation (IP) to the new designation of DATA at a 1:1 ratio, which will automatically transition for existing holders.

The rebranding effort includes the launch of Trace, an on-chain registry for AI training data, and an integration with Kled, a human data marketplace which allows opt-in participation. Data Foundation CEO Andrea Muttoni indicated that the decision to pivot was informed by insights gained in 2025, particularly through an AI data processing project that secured $15 million from venture capital firm a16z. He shared, "The answer became unambiguous: The form of IP pulling hardest was AI training data. Labs have effectively run out of internet to scrape." This highlights the evolving landscape of data sourcing where costs and legal risks have become significant challenges.

Muttoni articulated a vision for the new Data Network as a solution that integrates data processing, sourcing, and verification into a single platform, stating that currently, no company adequately addresses all three aspects. He emphasized, “That’s the gap we have been positioned to fill since day one, and that is the problem we are now dedicating our lives to solving.” As Data Network, the initiative aims to track over 1 billion data records with the mission of being the trust layer for AI training data.

Despite recent trading fluctuations, where the former token formerly known as IP saw its price drop significantly, the transition to DATA has already prompted a 14% increase in value, though it remains under $0.35, far below its previous high of $14.78.

On June 25, 2026, California Governor Gavin Newsom unveiled the country’s first AI-Unemployment Tracker, a public dashboard aimed at monitoring the impact of artificial intelligence on job losses in the state. This initiative responds to increasing concerns from labor experts and AI developers regarding the potential for automation to significantly disrupt the workforce.

The tracker, developed by the California Employment Development Department in collaboration with researchers at the California Policy Lab’s UCLA site, will update monthly. It will specifically focus on unemployment claims in occupations deemed highly exposed to AI technologies. State officials assert that the dashboard will allow policymakers to better assist workers who may be at risk of displacement, offering retraining programs, job-search help, and health coverage guidance.

Despite the dashboard's launch, state officials indicate that there is currently no evidence of widespread layoffs driven by AI. However, they acknowledge that certain groups of workers may be facing early signs of displacement. Till von Wachter, a Professor of Economics at UCLA, emphasized the importance of this initiative, stating, "This new tracker helps replace speculation with evidence, giving us a clearer understanding of what’s changing and how to best support affected workers."

This action is part of California's broader aim to actively shape AI policy, with Governor Newsom positioning the state as a proactive leader in addressing the implications of AI in society. The introduction of this tool reflects a growing acknowledgment among lawmakers, such as Senator Bernie Sanders and others, of the pressing need to address potential job disruptions caused by rapid advancements in artificial intelligence technology.

Moving forward, observers should monitor data emerging from the tracker and assess the implications of AI on the job market in California and beyond. As AI technology evolves, it will be critical to evaluate not only how these changes affect employment trends but also the effectiveness of the support measures being implemented for displaced workers.

Nous Research recently introduced animated mascot pets for its self-improving AI platform, Hermes Agent. This feature allows users to choose from over 3,200 animated sprites that reflect the agent's activities through various visual states.

The six states include:

• Idle: The pet remains still when the agent is inactive.
• Running: The pet indicates when a tool is executing.
• Thinking: The pet represents the agent when it is processing information.
• Waving: A sign that the agent has completed a task.
• Celebrating: The pet jumps upon the successful completion of a full plan.
• Failing: The pet shows a defeated state whenever the agent encounters an error.

The pets do not affect the agent's performance, caching, or token management, which was a key clarification in the feature's documentation. Despite their lack of utility in a traditional sense, the addition aims to enhance user interaction and make the Hermes platform more visually engaging.

The implementation of these pets operates across Hermes' Command Line Interface (CLI), Text User Interface (TUI), and desktop applications. Users can integrate their selected pet with a simple install command and can switch between mascots using specific commands. The integration aims to make terminal use more appealing and personalized.

With the pets feature disabled by default, users must actively choose to enable it, which is considered a thoughtful design choice. The small animated sprites add a layer of personality to the AI without complicating its function. This approach reflects a growing trend in software development that incorporates elements of engagement and fun into otherwise utilitarian tools.

A leaker has reported finding new strings in Claude Code version 2.1.190 that indicate Anthropic may bundle Fable 5 usage into weekly subscription plans. According to the leaked information, users will no longer need to purchase Fable 5 separately, as the updated code includes the phrase, "You’ve used your Fable 5 usage for this week." This indicates that usage limits will reset weekly, akin to a mobile data plan.

The leaked strings suggest a restructuring of Anthropic’s pricing model. Previously, after an initial free usage window, customers were expected to purchase usage credits separately to continue using Fable 5. However, the latest strings eliminate references to separate purchases, indicating a more integrated approach within subscription limits.

The change may be significant for users who previously faced restrictions due to the Donald Trump administration's ban shortly after Fable 5's launch, citing national security concerns over its potential misuse. The new subscription model might alleviate the limitations and encourage broader access to Fable 5.

Decrypt has verified the authenticity of the leaked strings through a direct download from npm of the macOS version of Claude Code. The alterations signify a concrete shift in how Anthropic is positioning Fable 5 in the marketplace.

What to Watch

• Anticipate further details from Anthropic regarding the new subscription model.
• Monitor any developments in the regulatory landscape, given the historical concerns surrounding the technology.
• Watch for user feedback on the new usage structure once it is implemented.

On June 24, 2026, physicist Henry Legg published a critique disputing Microsoft’s claims about its Majorana 2 quantum chip. According to Legg, Microsoft has not effectively demonstrated the existence of a topological qubit, which is crucial for the reliability of quantum computing. He outlined several issues in Microsoft's research, including faulty procedures and potential misinterpretation of experimental data.

Microsoft asserted that the Majorana 2 chip, claimed to be 1,000 times more reliable than its predecessor, signifies a significant breakthrough in moving toward practical quantum computing by 2029. However, Legg argues that the signals Microsoft associates with its topological qubit may simply be noise rather than proof of the device’s functionality. His commentary highlights the challenges in verifying topological qubits, stating that “the detection of a topographical superconducting phase is notoriously difficult because trivial states can mimic the signatures expected from a topological superconductor.”

In response to the criticism, Microsoft defended its findings, reiterating that AI technology has played a pivotal role in developing Majorana 2 by streamlining material selection and manufacturing processes. The company claims its approach could enhance the stability and reliability of quantum computers, as evidenced by the average stability of quantum information at 20 seconds and some qubits reportedly lasting up to one minute.

This development raises important questions about the future of quantum computing technologies and the validity of claims made by leading tech firms. Observers in the scientific community and the tech industry will likely scrutinize both Microsoft’s defense and Legg’s critique as the field advances.

What Happened

On June 24, 2026, OpenAI announced the launch of Jalapeño, its inaugural custom AI chip designed for large language model inference. Developed with Broadcom, this chip is part of OpenAI's strategy to build out its hardware capabilities, enabling the company to support applications such as ChatGPT more efficiently. Early versions of Jabapeño are reported to be in testing within OpenAI’s labs, and the company anticipates this chip will outperform current AI chips by delivering more computational power with lower energy consumption.

Why It Matters

This announcement marks a significant shift for OpenAI by moving towards a more self-sufficient hardware strategy. Previously reliant on Nvidia’s chips, Jalapeño is designed to enhance the processing capabilities needed for AI models while reportedly consuming less energy. OpenAI President Greg Brockman stated, "Jalapeño is part of our long-term, full-stack infrastructure strategy to make compute more abundant, resulting in AI which is faster, more reliable, more affordable for people and businesses, and can be used to solve more important problems." This shift could allow OpenAI to push the boundaries of AI more effectively, potentially leading to more widespread access and diverse applications.

What to Watch

Moving forward, it will be important to monitor the performance benchmarks of Jalapeño when they become available. The deployment of this chip in data centers is expected later this year, alongside the development of future generations that may support more extensive computing power. Additionally, the performance of Jalapeño in emerging AI applications, including its use with evolving models like GPT-5.3, will be key indicators of OpenAI's success in its hardware endeavors.

Qwable, a free local AI model replicating Anthropic's Fable 5 reasoning style, has been officially launched. This model is a fine-tuned adaptation of Alibaba's Qwen3.6-27B, built specifically to emulate Fable 5’s structured thinking in an accessible format, capable of running on standard consumer devices without additional costs.

Developed by a user known as Mia on Hugging Face, Qwable integrates techniques known as instruction fine-tuning, which allow it to generate outputs that mirror the deliberate, step-by-step reasoning style of Fable 5. The project aims to democratize access to advanced AI capabilities by removing the dependencies on costly APIs and restrictive policies.

Recent events in the AI community have underscored the urgency and relevance of such models. Last week, Anthropic faced backlash after their Fable 5 model was implicated in a controversy regarding invisible safeguards. Following these developments, a directive from the U.S. government prompted the withdrawal of this model for foreign nationals due to a jailbreak incident. In this context, the emergence of Qwable could provide a viable alternative for users seeking similar capabilities without the associated challenges.

Additionally, Qwable boasts a parameter count of 27 billion, placing it within the realm of high-capability models. These parameters determine the breadth of knowledge the model can utilize, with more parameters generally indicating enhanced capabilities. This local model allows users the freedom to execute AI tasks without financial constraints, marking a significant shift towards more inclusive AI solutions.

Moving forward, it will be interesting to observe how Qwable impacts the AI landscape, especially in light of competitors and community reactions. There are open questions regarding the potential for further developments and improvements to such models, particularly in areas of user experience and functionalities.

On June 22, OpenAI officially launched its GPT-5.5-Cyber model, part of the Daybreak cyber defense initiative, which aims to enhance the identification and resolution of software vulnerabilities. This model achieved 85.6% in the CyberGym benchmark, which assesses AI agents’ ability to replicate known vulnerabilities, surpassing Anthropic’s Mythos 5, which scored 83.8%.

The release of GPT-5.5-Cyber comes amid significant scrutiny of other AI models in the market. Anthropic had its Mythos 5 and Fable 5 models taken offline on June 12 due to a national security export control directive issued by the Trump administration, citing concerns about a "jailbreak" vulnerability that could allow users to bypass safety measures. This has left Anthropic unable to operate their advanced models, while OpenAI moves forward with its cybersecurity offerings.

Anthropic’s CEO, Dario Amodei, had previously articulated concerns about the capabilities of Mythos, emphasizing the potential risks associated with its deployment. In an essay, he likened frontier AI models to aircraft that safety regulators could ground for failing audits. Now, with the U.S. government effectively grounding Mythos, the competitive landscape in AI-driven cybersecurity is evolving quickly.

OpenAI’s approach contrasts with Anthropic’s more restrictive deployment of AI technologies. The designated release of GPT-5.5-Cyber to trusted defenders marks a significant step in AI-driven cybersecurity, as it demonstrates OpenAI's agility in development despite regulatory challenges faced by competitors.<\/p>