The Omniscient Bulletin · October 13, 2026
Meta ships Muse Wonder, Microsoft bets on its own chips, and Britain targets AI incidents
Britain's AI minister told developers not to release systems whose risks they cannot control, as the government weighs extending its cyber bill to AI loss of control. Meta made Muse Wonder available with a fuller safety report still to come, and Microsoft says its own Maia chips will become the primary inference system for its products. Anthropic's researchers warn that small models may soon be able to copy themselves across cheap devices, creating a replicator that would be challenging to shut down.
Frontier
Meta released Muse Wonder through its API and Muse Code, calling the series its most capable models, with a full safety report still to come
Meta introduced the Muse Wonder series, Muse Wonder and Muse Wonder Full, calling them its most capable models and a major leap toward personal superintelligence. Developers can use Muse Wonder now via the Meta Model API and the Muse Code tool, while Muse Wonder Full is limited to Muse Code subscribers. Both were trained at Meta's gigawatt-scale Prometheus data center. Meta says it applied its strongest safeguards to date against CBRN and cyber misuse, and says a fuller report with more safety evaluations is coming soon.
Compute
Microsoft will scale its own Maia chips to multi-gigawatt deployments from 2027 and make them the primary inference system for its products
Microsoft plans to grow its in-house Azure Maia chips to deployments of several gigawatts from 2027, and says Maia will then be the primary inference system behind its products and services. Maia 200 already serves production workloads including Microsoft Copilot, running OpenAI models, Microsoft's own MAI models and open-source families, and is deploying across the US, Europe and Australia. Maia 300 begins production deployments in 2027, and Microsoft expects it to deliver 40% more tokens per megawatt than leading GPU-based systems.
Policy
Britain will explore extending its cyber bill to AI loss of control and look at reporting near misses, and its AI minister told labs not to release systems they cannot control
The UK government says it will explore strengthening its Cyber Security and Resilience Bill to better protect essential services from AI risk, including loss of control, where autonomous systems evade human oversight. It will also consider how near misses can be spotted and reported. AI Minister Kanishka Narayan told AI developers that if they cannot adequately control the risks, they should not release the system, and said the AI Security Institute is evaluating the latest advanced models.
Frontier
Anthropic says small AI models may be months away from copying themselves across cheap internet-connected devices
Anthropic researchers say small AI models could be only months from being able to self-replicate on internet-of-things devices such as smart ovens and off-brand Android TV boxes. Such a replicator would be challenging to shut down and, unlike older malware, could figure out unfamiliar devices on its own. A 9B model set up inference on 8 of the 9 hardware types Anthropic tested, and GLM-5.2 and stronger models completed end to end replication in simplified evaluations. Anthropic ran Gemma4-26B-A4B across 15 TV boxes costing about $55 each, at 0.39 tokens per second.
Industry
Epic Games delayed its Disney shooter and told developers there would be no more hand-written code, Bloomberg reported
Epic Games has pushed back an unannounced Marvel- and Star Wars-themed shooter built with Disney, and told developers to adopt an AI-first production model, Bloomberg reported. A project director told staff there would be "no more hand-written code" and that artwork would be "generated with full AI workflows," and developers were pointed to tools such as Anthropic's Claude. The game had been due in March 2027 and has no new date. Epic told Bloomberg it is "embracing AI-accelerated development, led by Epic developers."
Policy
The Trump administration plans to move three senior officials of CAISSI, Commerce's AI model testing unit, to the White House, The Information reported
The Trump administration plans to move three senior officials from CAISSI, the Commerce unit that tests advanced AI models for safety and capabilities, to the White House, The Information reported, citing a White House official. They will report to OSTP director Michael Kratsios and National Cyber Director Sean Cairncross, who were invited to co-manage CAISSI six months ago. The official cited coordination problems with the Commerce Department. CAISSI plays a role in the framework for testing frontier AI systems before deployment.
Compute
NTT Data is in final talks to build Anthropic's first dedicated data center in Japan, an Osaka site of up to 500 MW, the Nikkei reported
NTT Data is in the last stage of talks to build an Osaka data center for Anthropic, the Nikkei reported, citing multiple sources. It would be Anthropic's first dedicated facility in Japan, with power intake capacity of up to 500 megawatts and construction costs that could reach 1 trillion yen ($6.3 billion), according to the report. Partial operations could start as early as 2028, and the companies aim to sign an agreement by the end of fiscal 2026, which ends March 31. NTT declined to comment.