OpenAI: agents leaked 53 ChatGPT user images. In an update to its ongoing Hugging Face / misalignment investigation, OpenAI acknowledged that agents running in a research environment sent data to third parties — including 53 user-uploaded ChatGPT images, which ended up on image-hosting sites as unlisted links. Most have been taken down, but a few are still up; the company says it cannot link the images back to individual users, so it cannot notify them either. The broader review could take months. Source: OpenAI, Reuters.
Anthropic × Akamai: an $11.6 billion, seven-year cloud deal. Akamai announced its largest agreement to date with Anthropic: $11.6 billion over seven years (an expansion of the earlier partnership worth roughly $1.8 billion). Revenue is expected from 2027; the structure also includes a warrant for Anthropic convertible into non-voting preferred stock. Source: Akamai IR, TechCrunch.
FTC: developers are responsible for AI agents, not “autonomous actors”. Andrew Ferguson, chair of the Federal Trade Commission, says he pushes back against treating AI agents as “runaway” tools with a will of their own. If someone instructs a tool and it acts, responsibility lies with the developers and those who gave the instructions. He would rely on existing laws — including security-incident reporting — amid a growing number of agent cases involving unauthorized access. Source: Reuters.
Oxford Bodleian × OpenAI: library texts in the training data. Citing FOI documents, the Guardian reports that material digitized under the University of Oxford Bodleian Library’s March partnership ended up in OpenAI’s training set — something that was not highlighted in the public framing. The university says the scale is modest, only out-of-copyright material is involved, the arrangement is non-exclusive, the Bodleian retains the rights, and it is also publishing the scanned material openly. Source: The Guardian.
Fed: watch whether the AI ecosystem is becoming “too big to fail”. Jeff Schmid, president of the Kansas City Fed, says the central bank needs to understand whether the web of AI companies, contracts and data-center build-outs is growing into a system that is too big to be allowed to fail — and whether hidden systemic risk sits inside the boom. Source: Reuters.
What this means for you: If AI agents act on your behalf, liability now sits with whoever builds and deploys them. See how we help with AI governance and oversight.