AI Giants Warn Congress: Please Regulate Us (Very Gently)

In what can only be described as the most elegant regulatory capture play since private equity discovered carried interest, OpenAI and Anthropic have begun briefing House Homeland Security Committee staff on their "cyber-capable AI models." Translation: two companies with a vested interest in minimal oversight are privately educating the people who will eventually regulate them about dangers only they seem equipped to manage.

The briefings were conducted behind closed doors, naturally. Nothing says "transparent responsible disclosure" quite like secret meetings with lawmakers about existential threats. The fact that these briefings represent "one of the first" such congressional sessions with AI giants is the real tell—not because it's early engagement with policymakers, but because it means OpenAI and Anthropic are writing the first draft of their own regulatory framework, pen in hand, while regulators take notes like undergrads at an optional lecture.

This is the oldest playbook in Silicon Valley: identify an emerging regulatory threat, position yourself as both the problem and the only credible solution, brief sympathetic lawmakers in private, and emerge as the reasonable party that "called for regulation" all along. Anthropic's constitutional AI and OpenAI's safety theater get sold not as marketing but as civic responsibility. Lawmakers feel informed. Founders feel virtuous. Competitors get shut out of the room.

In three years, when Congress passes watered-down AI legislation with exemptions specifically tailored to these two companies' business models, everyone involved will point back to these briefings as evidence of good faith. Regulatory capture doesn't announce itself—it presents itself as a briefing on cyber threats.