White House “Super Intelligence” accord: a voluntary safety framework for the big labs — and a new name in federal communications. On Tuesday at the White House, Donald Trump and Sundar Pichai (Google), Dario Amodei (Anthropic), Mark Zuckerberg (Meta), Greg Brockman (OpenAI), Elon Musk (xAI / SpaceXAI) and Jensen Huang (Nvidia) signed the White House Accord on Super Intelligence / Joint Commitment on Frontier Responsibilities. Its four voluntary layers: internal controls to monitor capabilities and alignment (including cyber, biological and chemical threats), a responsible internal team, an independent external auditor, and an independent committee of the board. The accord is not a legal obligation and carries no penalties. In parallel, Trump signed an executive order directing the executive branch to use the terms “Super Intelligence” / “SI” instead of “Artificial Intelligence” / “AI” wherever legally permitted. Source: White House — executive order, fact sheet, Washington Examiner (full text), UPI.
FTC opens a probe into OpenAI, Anthropic and other AI companies over product risks. A Federal Trade Commission spokesperson confirmed to CNBC that the agency has opened an investigation into OpenAI, Anthropic and other artificial intelligence companies over the potential dangers posed by their products. The New York Post first reported the move; the spokesperson declined to name the other companies involved. The probe comes amid mounting scrutiny of the labs’ safety practices — following earlier agent incidents and public safety debates. Source: CNBC, New York Post.
OpenAI disrupts a coordinated model-distillation campaign — and ties its core to Moonshot AI (Kimi). OpenAI says that from July 1 it identified coordinated activity attempting to extract its models’ protected internal reasoning. This is so-called adversarial distillation: extracted reasoning can be used to train another model or copy its capabilities. Activity peaked on July 24–25 (16,000 requests from more than 4,000 users); a related pattern appeared across a cluster of more than 15,000 users, which was fully disrupted by July 28. According to the company, the operators did not break its encryption or access stored conversations; OpenAI attributes the core of the activity to individuals associated with Moonshot AI, the developer of Kimi. The information was also shared with industry partners through the Frontier Model Forum. Source: OpenAI.
Google DeepMind: SynthID Bio — watermarks for AI-designed proteins, a Nature paper and open tools. The biological version of the SynthID family embeds an imperceptible, verifiable signal in AI-generated protein sequences and predicted 3D structures — the aim is to protect biosecurity and the integrity of scientific databases. In wet-lab tests on three target proteins (VEGF-A, the receptor-binding domain (RBD) of the SARS-CoV-2 spike protein, and PD-L1), watermarked binders matched the hit rate and binding strength of non-watermarked ones. The method is published in Nature, and the code and data are being opened up to the research community. Source: Google DeepMind, Nature.
Gemini: Skills arrive — and replace Gems. Google is rolling out skills in Gemini chat worldwide: saved instruction sets you can invoke with a forward slash (/) in the prompt box — several at once if needed, with reference files. Skills replace Gems, and existing Gems will migrate automatically. Support for Gems ends starting in November for personal accounts, in March 2027 for Workspace business, enterprise and nonprofit customers, and in June 2027 for education customers. Source: Google Keyword.
Anthropic: open-weight GLM-5.3 approaches Mythos Preview’s cyber capabilities — with weak safeguards. According to Anthropic’s Frontier Red Team, Zhipu AI / Z.ai’s GLM-5.3 can autonomously build end-to-end exploits — working code that takes advantage of vulnerabilities: on ExploitBench it succeeded in 50 of 410 attempts (Claude Mythos Preview: 56/410). In simulated tests, its safeguards could be bypassed 64–100% of the time with simple techniques. The US NIST CAISI calls it the most cyber-capable open-weight model released to date, lagging the US frontier by about four months overall. Source: Anthropic, NIST CAISI.
What this means for you: A single day shows the political framing (a voluntary accord plus an FTC probe), the labs’ defensive reflexes (countering distillation, biosecurity watermarking) and product-side automation (Gemini Skills). If your company is rolling out AI, clarify accountability, oversight and regulatory risk in advance — see our AI governance services, or book a free 60-minute AI audit.