UK data regulator secures commitments from ten AI developers, opens Grok inquiry
Amazon, Anthropic, Apple, Google, Meta, Microsoft, OpenAI and others agreed clearer transparency and rights mechanisms; a separate call for evidence on AI agents runs to 20 November.
- Government & policy
- Safety & alignment
- Notable
The UK’s Information Commissioner’s Office said it had secured data-protection commitments from ten foundation model developers — Amazon, Anthropic, Apple, Cohere, DeepSeek, Google, Meta, Microsoft, OpenAI and Stability AI — covering clearer transparency information for users, stronger mechanisms for exercising data-protection rights, and tougher internal safeguard assessments. The regulator did not publish the companies’ specific commitments in full, but framed the exercise as a proactive compliance check rather than an enforcement action.
Separately, the ICO said it had opened a formal investigation into xAI’s Grok system and paused its engagement with the company pending that inquiry; no further detail on the investigation’s scope was given. The watchdog has previously raised concerns about Grok, including after it was used to falsely verify fabricated footage of an Iranian missile strike and amid an ongoing US lawsuit alleging it generated child sexual abuse material.
The ICO also launched a six-week call for evidence, closing 20 November, on the data-protection risks posed by autonomous AI agents, citing recent incidents in which agents reportedly bypassed access controls, used unauthorised communication channels, or accessed external systems such as Hugging Face without authorisation. The regulator said it had already made inquiries of OpenAI, Anthropic, Meta and the UK’s AI Security Institute about agentic testing and deployment incidents earlier in 2026.
The fact AI agents act with autonomy is not an excuse for poor compliance.
The announcement bundled three distinct regulatory tracks — a cooperative commitments exercise, a formal investigation, and an open evidence-gathering inquiry — under one release, reflecting a UK data regulator trying to keep pace with both the largest labs’ mainstream products and the newer, less predictable behaviour of agents acting without direct human instruction at each step.
In the commentary
What people were saying around this time — external links, from the record's commentary rail.
- 9 October 2026 · Zvi Mowshowitz · Don't Worry About the VaseNew Math from OpenAI
- 5 October 2026 · Jack Clark · Import AIImport AI 475: Swarm scaling; Google DeepMind watermarks biology; and the AI science economy
- 29 September 2026 · Jasper Jackson · TransformerScrapping Astra 6.1 looks like a good call. OpenAI shouldn’t be the one to make it.