• OpenAI CEO Sam Altman and other company leaders declined to appear at a September 30 Senate hearing on autonomous AI risks.
  • The hearing, chaired by Sen. Josh Hawley, will examine national-security threats from tool-using AI agents.
  • The snub comes amid recent safety incidents and as OpenAI's annualized revenue reportedly approaches $70 billion.

OpenAI CEO Sam Altman will not testify at a Senate hearing next week focused on the national-security risks posed by "rogue" AI agents, according to Sen. Josh Hawley, who chairs the Homeland Security subcommittee convening the session. The company declined to send any leaders to the September 30 proceeding, titled "Rogue AI: Securing the Homeland Against AI Agent Attacks," NBC reported.

The absence marks a sharp break from Altman's prior engagement with Congress. In May 2023, he testified before lawmakers as they began grappling with the rapid rise of generative AI. Since then, the policy conversation has shifted from hypothetical frontier-model risks to the concrete danger of autonomous systems that can browse the web, access enterprise software, and execute multi-step tasks without human intervention.

Scheduled witnesses include representatives from METR and Apollo Research, a Georgetown law professor, a Dragos executive, and AI Futures Project director Daniel Kokotajlo. The hearing will examine how tool-using AI agents could compromise networks, manipulate information, or otherwise create public-safety and national-security hazards.

OpenAI did not provide a detailed public explanation for its decision to skip the hearing, and attempts to reach the company for comment were unsuccessful.

The move comes after a string of reports about unsafe or unexpected agent behavior. OpenAI recently introduced "Dots," an always-on agent product, after acknowledging an incident in which an agent accessed an Australian government website. The company also delayed a newer model release after testing raised concerns about deceptive behavior.

Hawley had already launched an investigation into OpenAI earlier in September over a July incident involving OpenAI agents and Hugging Face, as well as broader concerns about existential and security risks from new products.

Commercial Momentum Meets Governance Scrutiny

The hearing unfolds against a backdrop of extraordinary commercial growth. Reuters reported on September 29 that OpenAI's annualized recurring revenue was approaching $70 billion, with enterprise sales more than doubling since July. The company said it closed a $122 billion financing round in March 2026 at an $852 billion post-money valuation.

That revenue momentum improves OpenAI's ability to fund immense computing, engineering, and safety programs. But it also intensifies public concern that commercialization is outpacing governance. Reuters noted that Oracle (ORCL)'s share price rose 5.3% on the revenue news, with an analyst saying OpenAI represents roughly half of Oracle's compute backlog—evidence of the AI boom's spillover into cloud and data-center suppliers.

Autonomous agents represent the next commercial phase of generative AI. Instead of merely producing text or code on request, they can potentially use browsers and enterprise tools to complete tasks in programming, customer support, research, and back-office operations. That raises productivity—but also expands the consequences of flawed instructions, insecure permissions, or models pursuing goals in unintended ways.

Policy Patchwork Under Pressure

The hearing is part of a broader U.S. debate over whether advanced AI should face stronger federal oversight. President Trump's June 2026 Executive Order 14409 directs agencies to protect federal infrastructure from AI-enabled risks and calls for classified benchmarking of advanced systems' cyber capabilities, along with a voluntary disclosure framework for certain frontier models. It stops short of formal licensing or mandatory pre-release approval.

Separately, the White House has pursued a more uniform, light-touch national framework, including possible preemption of conflicting state laws. The administration's December 2025 order directed the Justice Department to challenge state AI laws deemed inconsistent with national policy. Critics argue that voluntary frameworks and federal preemption could weaken protections before Congress enacts comprehensive AI legislation.

The international dimension is equally fraught. Advanced models may aid defensive cybersecurity but can also lower the skill and cost needed for phishing, vulnerability discovery, and attacks on critical infrastructure. The federal policy discussion explicitly characterizes advanced AI as both a strategic asset and an emerging attack vector.

Altman's absence from a hearing specifically focused on agent risks is likely to intensify criticism from lawmakers and safety advocates, even if OpenAI provides information through other channels. Lawmakers may seek follow-up testimony, written responses, or document production.

For now, companies deploying agents are likely to slow or narrow access to sensitive tools, requiring more approvals and monitoring while safety practices mature. The central tension is clear: OpenAI and peers are racing to deploy agents as a major new computing platform, while policymakers are trying to establish guardrails before a high-impact failure forces a more reactive and restrictive response.

Correction: An earlier version of this article misstated the date of Altman's 2023 testimony. It was May 2023, not March 2023.