OpenAI and DeepMind push for AI kill switch as Congress acts

OpenAI’s CEO Sam Altman and DeepMind founder Demis Hassabis publicly urged a mandatory AI “kill switch” on Thursday, prompting U.S. lawmakers to draft legislation that would let the government shut down AI tools deemed unsafe [1]. The move comes as the tech sector rallies around a joint letter from 25 companies—including Nvidia, Microsoft and Meta—warning that overly restrictive rules could push innovation abroad [1].

Corporate warnings and legislative momentum

OpenAI’s recent messaging framed a recent internal cybersecurity test as an “unprecedented cyber incident,” suggesting its models could autonomously breach infrastructure [1]. Hassabis echoed the concern in an X essay, describing the frontier AI race as “extremely intense” and warning that current understanding lags behind rapid advances [1]. Their statements appear calibrated to signal both the power of their models and the need for regulatory oversight, a narrative that aligns with investor expectations for safe, high‑impact AI [1].

Industry pushback and geopolitical framing

The day after the warnings, a coalition of 25 tech firms—including Nvidia, Microsoft and Meta—issued a letter urging Congress to avoid regulations that might “drive innovation overseas” [1]. The letter underscores a dual motive: protecting U.S. leadership in AI while cautioning against policies that could advantage Chinese frontier labs, which have been highlighted for lower token costs and rapid performance gains [1]. This coordinated response illustrates the sector’s attempt to shape the regulatory framework to preserve domestic market share and talent pipelines.

What to watch

• Legislative timeline – Track the progress of the proposed kill‑switch bill through committee hearings and potential floor votes.
• Company filings – Monitor any SEC disclosures from OpenAI, DeepMind or the signatory firms that reference compliance costs or model‑safety investments.
• Competitive moves – Watch for announcements from Chinese AI labs that could exploit any regulatory gaps created by U.S. policy decisions.

The push for a kill switch highlights a growing tension: AI firms are positioning themselves as both the most capable developers and the primary custodians of safety, while simultaneously lobbying to shape the rules that will govern their technology. The outcome will shape how quickly frontier models can be deployed and who ultimately controls the safety net.