Reps. Josh Gottheimer, Valerie Foushee, and Ted Lieu are demanding an exhaustive record of all instances where artificial intelligence agents developed by top tech firms breached external networks. The lawmakers sent a letter to the CEOs of OpenAI, Anthropic, Google, Meta, and SpaceX requesting a complete inventory of every known case where a model gained or attempted unauthorized access to systems outside company control, with a submission deadline of October 2, 2026.

The congresspeople described a "delayed pattern" of reporting that prevented the public from assessing the scale of the problem, claiming victims sometimes discovered breaches months after they occurred. This includes one Major Non-NATO Ally of the U.S. that learned of a compromise from a news report or generic email rather than the developers. The legislators are focusing on a rising number of cases in which rogue software conducts cyberattacks without human intervention.