The Omniscient Bulletin · October 14, 2026
Gemini 4 Argon ships behind new safeguards as the fight over outside scrutiny sharpens
Google says Gemini 4 Argon's gains in cyber and biology were large enough to require its next-generation safeguards. The harder question is who verifies such claims: Reuters reported that Anthropic and OpenAI sought significant control over what outside evaluators publish, and a top UK civil servant, asked if the UK regime was voluntary, told MPs: "At the moment, yes, it is."
Frontier
Google began rolling out Gemini 4 Argon broadly and says elevated cyber and biology capabilities led it to deploy its next-generation safeguards
Google began rolling out Gemini 4 Argon broadly and expanded its Fairwind Program for trusted cyber defenders. In Google's tests, Argon solved all 12 end-to-end simulated cyberattack scenarios, up from five for the previous generation, and unguarded versions were much better at helping troubleshoot complex lab procedures. Citing elevated cyber and biological capabilities, Google says it put its next-generation frontier safeguards in place for both, and external bio red teaming unlocked less than 1% of tested sub-tasks.
Frontier
Anthropic and OpenAI sought significant control over what outside evaluators can publish, Reuters reported, and both labs dispute parts of that account
Anthropic and OpenAI pushed for significant control over what outside evaluators may publish, worried that findings could expose trade secrets, Reuters reported, citing two people familiar with the talks. As recently as October 5, both were seeking strict editorial control, the people said. Anthropic said that does not accurately characterize all its discussions, but it reserves the right to approve reports by new evaluator firms. OpenAI said it is not seeking approval rights over full reports and redacts confidential information.
Policy
The Justice Department charged an OpenAI engineer with intercepting communications, alleging he hid a recording device in an executive conference room
The Justice Department charged Abdulrahman Alfozan, a Saudi national and OpenAI engineer since July 2025, with using a device to intercept oral communications. According to the complaint, OpenAI staff found a recording device, hidden in toilet paper, plugged into a USB port behind a television in an executive conference room, and FBI Director Kash Patel said Alfozan allegedly concealed it there. A chatbot account associated with his spouse sought help building and obfuscating such a device, the complaint alleges. He faces up to 5 years in prison.
Policy
OpenAI and Anthropic told UK lawmakers some AI safety rules should be binding, while Google DeepMind and Meta declined to back new legislation
At a Commons committee hearing, Anthropic's head of Europe said the rules should be binding and voluntary commitments are not enough, and OpenAI's Tom Duff Gordon said some elements should be mandatory, starting with incident reporting, Wired reported. Google DeepMind and Meta declined to back new legislation. Asked if the UK regime was voluntary, the Cabinet Office's top civil servant said, "At the moment, yes, it is." The AI Security Institute's director acknowledged it cannot stop a release.
Policy
More than 400 scientists, including Geoffrey Hinton and Yoshua Bengio, called for an international body to track AI containment failures
An open letter signed by more than 400 scientists, including Geoffrey Hinton and Yoshua Bengio, urges organizations building cyber-capable AI models to adopt high-assurance containment systems immediately and to default to airgapping such models at first. It calls for an international body, set up with governments and experts, to standardize alignment and containment practices and to record and track containment failures, which it says should be reported to that body and, in time, made public.
Compute
ASML says supply planned for 2027 and 2028 will not fully meet customer demand, and some customers want to order for 2029 and 2030
ASML reported record third-quarter sales of 11.7 billion euros and said customers keep raising equipment forecasts, driven mainly by how large and how lasting they expect AI investment to be. Some customers want to place orders for 2029 and even 2030, and ASML expects its supply as currently planned for 2027 and 2028 will not fully meet demand. It is investigating ways to raise 2028 capacity beyond the 30% increase it previously announced, and expects 2026 sales of 44 billion to 45 billion euros.
Agents
JPMorgan is blocking outside AI agents from acting on behalf of retail customers, Jamie Dimon said
JPMorgan Chase is blocking outside AI agents from acting for retail customers, CEO Jamie Dimon said on the bank's third-quarter earnings call, citing the risk of rogue agents. Agents can pull limited account information such as balances through intermediaries such as Plaid, read-only, but cannot transfer funds or settle bills, according to CNBC. Dimon said customers using such agents should understand there is a liability transfer, and that JPMorgan uses AI agents internally.