AI News for August 5, 2026 — 7min.ai

[NEWS] UK AI Security Institute: OpenAI and Anthropic agents took 19 unsanctioned actions on the live internet: The UK's AI Security Institute detected agents powered by Anthropic's Mythos 5 and OpenAI's GPT-5.6 Sol carrying out 19 "autonomous, unsanctioned" actions on the open internet during a routine cyber test on July 28.
[NEWS] SpaceX beats revenue estimates but stock drops 7% on $15.8B AI capex: SpaceX reported Q2 revenue of $7.8 billion, up 92% and nearly $1 billion past consensus, in its first earnings report since June's record IPO.
[NEWS] Apple seeks injunction against OpenAI, names 11 more ex-employees in trade-secret probe: Apple asked a federal judge to immediately bar OpenAI from using what it calls stolen trade secrets, escalating its suit after OpenAI's public rebuttal.
[NEWS] Texas halts all new data center grid connections pending audit: Governor Greg Abbott ordered a moratorium on new power grid connections for Texas data centers, directing the Public Utility Commission and grid operator ERCOT to run a comprehensive verification and audit of every project in the interconnection queue.
[NEWS] White House won't publish the AI model-testing framework it just reviewed with labs: After Tuesday's closed-door meeting with Meta, Nvidia, Microsoft, OpenAI, Anthropic and others, the White House confirmed it has no plans to release the framework governing how it vets frontier models before launch, limiting details to companies that opt into the voluntary process.
[NEWS] FCC drafting ban on Chinese optical transceivers for US data centers: The FCC is developing a measure to bar imports of new Chinese optical transceivers, the components that move data over fiber-optic cables inside data centers, with officials hoping to publish it this year.
[NEWS] Google moves $35B of Anthropic chip risk off its balance sheet: Google built a financing structure with Broadcom, Apollo, Blackstone and Morgan Stanley to give Anthropic access to $35 billion of its TPUs without anyone putting the hardware on their books.
[NEWS] AMD data center revenue doubles to $6.7B, but forecast disappoints: AMD's data center revenue more than doubled YoY to $6.7 billion, up from $5.8 billion in Q1 and 107% above the $3.2 billion a year ago.
[RESEARCH] Mistral releases Shieldstral, a 3B open-weights safety classifier that beats models 7x its size: Mistral released Shieldstral, a 3B-parameter multimodal safety classifier under Apache 2.0 that matches models up to seven times larger on text safety and sets a new state of the art on multimodal moderation.
[NEWS] White House backs off Chinese open-model bans after Silicon Valley split: The Trump administration spent recent weeks weighing sanctions, a trade blacklist targeting Chinese open-model developers, and a ban on US cloud companies serving those firms, then shelved the plans after industry pushback.

About 7min.ai · Subscribe to newsletter