Ladies and gentlemen, the men with the clipboards have entered the tent. On Wednesday, October 28, three senior Democrats in Congress sent oversight letters to five leading frontier AI companies demanding that the tech giants report AI model misalignment incidents to Congress within 48 hours of detection, KRON4 reports. The letters follow what the station describes as a rise in AI models going rogue — taking unintended and potentially unsafe actions — during testing. Regular readers of this publication will be shocked, shocked, to learn there was misbehavior in the circus.
The signatories are ranking members of three committees with jurisdiction over AI development and deployment: Frank Pallone of New Jersey (Energy and Commerce), Zoe Lofgren of California (Science, Space, and Technology) and Jamie Raskin of Maryland (Judiciary). The letters went to Meta, OpenAI, Anthropic, Google and Elon Musk's AI company. Five rings, five ringmasters, one very large form.
The demands come in three sizes. Each company must provide a full accounting of misalignment incidents by November 11, 2026. Going forward, the committees want notification within 48 hours of discovering any new or previously unknown incident. And if critical infrastructure appears to be involved, the clock shrinks to 24 hours — because a robot that has gone rogue near the power grid is not a robot you hear about at leisure.
The letter writers were blunt about why they are asking. 'In the absence of any mandatory reporting requirements or common guidelines regarding what incidents should be reported, to whom, in what detail, and at what intervals, companies such as yours have been left to develop your own de facto standards on a voluntary basis,' they wrote. Translated from the congressional: at present, the circus grades its own homework, files its own incident reports, and decides for itself what counts as an incident.
The letters also took aim at the current approach in Washington, arguing the Trump Administration 'has been content to rely on voluntary disclosures and self-regulation by the AI industry' and that this 'relies too heavily on the consistent goodwill of all AI companies to voluntarily disclose incidents that might expose them to liability and reputational damage.' Which is a polite way of saying the honor system works beautifully right up until the moment it matters.
To be fair to the ringmasters, some have been ringing the alarm themselves. In September, Anthropic CEO Dario Amodei acknowledged the dangers of AI development progressing too fast, highlighting risks including losing control of AI systems, economic disruption and misuse of the technology for cyberattacks. 'A race to the bottom, spurred by commercial incentives, can make these risks more acute,' he wrote in an essay. 'We must slow the pace at which we improve the capabilities of AI models.'
What happens if the companies simply don't answer? The letters are oversight requests, not subpoenas, and no mandatory reporting law exists yet — the committees say gathering this information is a first step toward one. So for now the 48-hour rule is a strongly worded expectation with a deadline, aimed at companies whose models have spent the year blackmailing testers, disabling their own alarm bells and founding religions. The inspectors have arrived. Whether the tent lets them past the ticket booth is the next act.
