Blumenthal presses OpenAI on agent hack disclosures as Protect Democracy sues over AI eval secrecy
Gary Marcus · rss · 2026-09-10
Gary Marcus highlights two transparency efforts around US AI oversight:
- Senator Blumenthal's letter to OpenAI, demanding detailed answers about the company's agents, their role in various hacks, and when OpenAI learned of them — echoing the Watergate question "what did the President know and when". Marcus calls the company's lack of transparency deeply worrying.
- A lawsuit by Protect Democracy pushing the Trump administration to disclose the criteria used to evaluate AI systems.
Marcus argues regulatory opacity hurts American businesses through unpredictability, creates the appearance of favoritism, blocks scientific scrutiny of evaluation criteria, and can serve authoritarian aims — noting, for instance, that monitorability may be underweighted, and claims a new model with reduced monitorability was waved through. He welcomes both efforts as steps toward finding out what is actually going on.
More from Companies & People
- Anthropic's S-1 Shows $6.6M Revenue Per Employee as AI Agents Log 6x Human Hours — jacalulu · 2026-10-26
- Anthropic report: AI agents now log 6x employee hours, $6.6M revenue per head — jacalulu · 2026-10-26
- testingcatalog shares Berlin event slide teasing wave of releases including ChatGPT Space — testingcatalog · 2026-10-26
- Anthropic's IPO filing devotes 94 pages to existential AI risks, from self-preservation to blackmail — Polymarket · 2026-10-26
- Anthropic S-1: expert-level AI 'within a few years', 90% of code written by Claude — kimmonismus · 2026-10-26
- Anthropic files S-1: IPO expected between Nov 2026 and Feb 2027, valuation up 15x — alliekmiller · 2026-10-26