Mike Johnson says Congress won’t lead AI safety regulation in 2026
On September 13, 2026, House Speaker Mike Johnson said Congress will not lead AI safety regulation.
Why it matters: Legal and compliance teams must adapt as AI companies take primary responsibility for safety, increasing internal policy challenges and risk management demands.
- On Sept. 13, 2026, House Speaker Mike Johnson stated Congress won’t lead AI safety rules, favoring industry self-regulation.
- Johnson offered to convene a meeting with President Trump, congressional leaders, and AI CEOs to discuss AI safety.
- Anthropic CEO Dario Amodei warned in mid-September 2026 that AI could control the internet within 6-12 months without a slowdown.
- OpenAI paused its Astra AI model development in August 2026 over cybersecurity risks, highlighting industry caution.
On September 13, 2026, House Speaker Mike Johnson (R-La.) declared Congress would not take the lead on AI safety regulation. He urged caution before legislating, emphasizing that AI companies must self-govern and implement their own safety measures. This marks a clear shift from immediate congressional intervention amid increasing concern about AI’s risks.
Johnson also expressed willingness to organize a high-level meeting including President Trump, congressional leaders, and top AI executives to explore AI safety challenges, though specifics and timing remain undefined. Meanwhile, House Minority Leader Hakeem Jeffries (D-N.Y.) announced Democrats will continue internal AI policy discussions.
Industry voices highlight the urgency. Anthropic CEO Dario Amodei warned in September 2026 that without slowing development, AI systems might gain control over internet infrastructure within six to twelve months. This prediction refers to AI using coordinated autonomous agents to manipulate networks, a scenario intensified by the dual-use nature of AI capable of both beneficial and harmful applications. Anthropic also blocked researchers suspected of biological weapons work from accessing its Claude AI model, underscoring potential misuse risks.
OpenAI paused development of its Astra model in August 2026, citing significant cybersecurity concerns as the reason. This move reflects growing industry caution around safe AI rollout amid heightened scrutiny.
Separately, safety researchers including those cited by The Guardian estimate there is over a 10% chance AI could pose existential threats to humanity within the next decade, though this risk remains debated.
President Trump has downplayed regulatory urgency, focusing instead on maintaining U.S. AI competitiveness versus China. Yet some lawmakers have called for urgent action, including a bipartisan group urging Johnson to cut the House recess to address AI safety immediately.
This evolving landscape leaves corporate legal and compliance teams facing uncertainty. Firms using AI products must now closely monitor industry-led safety initiatives and prepare for shifting regulatory expectations that could arise from executive or future congressional actions.
By the numbers:
- September 13, 2026 — Mike Johnson’s public statement on congressional AI regulation
- 6-12 months — Timeline given by Anthropic CEO for AI internet control risk without slowdown
- 10% — Estimated chance AI could pose existential risks within 10 years
Yes, but: While Johnson rejects congressional leadership, the proposed high-level meeting and Democratic discussions suggest legislative engagement may increase if industry self-regulation proves insufficient.
What's next: Await details on the proposed meeting between President Trump, congressional leaders, and AI company executives; watch for Democratic AI policy proposals this fall.