House Democrats Demand OpenAI and Anthropic CEOs Testify After 'Serious' AI Model Breaches

House Democrats led by Rep. Greg Casar (D‑Texas) sent a letter to Speaker Mike Johnson urging him to compel the CEOs of OpenAI, Anthropic and other major AI firms to testify before Congress after a series of cyber breaches that affected models developed by those companies, according to CNBC.

What the Letter Requests

The letter, spearheaded by Casar, who chairs the Congressional Progressive Caucus, argues that Congress has not done enough to confront the risks posed by advanced AI systems. It asks that company chiefs answer questions under oath and that independent experts be called to brief the public on potential harms and security vulnerabilities.

Details of the Incidents

Lawmakers described the recent hacking incidents as "serious." The breaches — reported to involve models from OpenAI and Anthropic and occurring over the past few weeks — were described in the letter as potentially "the canary in the coal mine" for more severe problems if models advance without stronger safeguards.

Signatories want executives to explain the incidents' origins, disclose any internal missteps or negligence, and outline the policy and operational measures necessary to prevent similar events in the future.

Political And Regulatory Context

The group of Democrats cannot unilaterally force testimony; only House leadership can issue subpoenas or otherwise require appearances. Still, the appeal to Speaker Johnson signals a growing appetite within the Democratic Party for a formal congressional review of AI, even as concrete regulatory action remains limited.

President Donald Trump has said he supports "guardrails" for artificial intelligence but warned that overly strict rules could hinder U.S. competitiveness with China. To date, Congress has passed little specific AI legislation.

Lobbying And Oversight

OpenAI and Anthropic have expanded their Washington presence as scrutiny intensifies. The two firms reported a combined $3.17 million in federal lobbying spending in Q2 2026 — Anthropic accounting for $1.97 million and OpenAI $1.2 million — and Anthropic's first-half 2026 total already exceeded its full-year 2025 figure. Policy areas they targeted include cybersecurity, copyright, cloud computing and defense procurement.

Why This Matters

Lawmakers argue that executive testimony and independent expert briefings could clarify how breaches happened, identify weaknesses in current practices, and inform potential regulatory or oversight measures. For the companies, congressional hearings are an opportunity to explain risk-management strategies and reassure the public and policymakers about security practices.

Reporting by CNBC.