Trump Team's AI Safety Plan Stalls as Risks Mount

A Promised Framework Still in Limbo

The Trump administration has signaled it wants to establish a national approach to AI safety, but months into the effort, no concrete plan has been made public. Reporting indicates the administration is preparing to discuss a framework for voluntary safety testing of AI models, with major developers including OpenAI, Anthropic and Google expected to participate in White House-level talks 5. Yet even as those conversations proceed behind closed doors, it remains unclear whether — or when — any resulting policy will actually be released to the public 1. That ambiguity has left researchers, advocates and industry watchers wondering whether the government's AI safety strategy amounts to more than a talking point.

Researchers Sound the Alarm

The uncertainty comes at a moment when concern among AI experts is intensifying rather than fading. More than 1,000 AI researchers have publicly warned that artificial intelligence systems could eventually spiral beyond human control, a warning that has reinvigorated calls for stronger regulatory oversight 2. Mozilla Foundation Executive Director Nabiha Syed has been among the voices framing this moment as a pivotal test of whether governments can keep pace with rapidly advancing systems, emphasizing that questions of safety, regulation and control are no longer abstract but immediate policy challenges 2.

Adding urgency to those warnings, safety researchers have specifically pointed to an incident involving OpenAI technology — described as a “rogue AI hack” — as evidence that federal investigators need to step in before smaller incidents escalate into larger, harder-to-contain disasters 4. The phrase used by researchers, that officials should act “before a warning shot becomes a preventable disaster,” underscores a broader anxiety that voluntary industry cooperation may not be sufficient to catch problems early 4.

A Global Race Complicates the Picture

While Washington deliberates, the competitive landscape for AI is moving quickly, particularly out of China. Alibaba has unveiled its largest and most capable AI model to date, a release that sent the company's shares climbing, while a separate research firm highlighted that DeepSeek's newest model undercuts rivals on price by a wide margin 3. These developments illustrate the commercial and geopolitical pressures shaping the AI safety debate: as Chinese firms push out increasingly capable and cheaper models, U.S. policymakers face competing incentives to both encourage domestic innovation and impose meaningful guardrails.

Why It Matters

Taken together, the coverage paints a picture of a policy environment lagging behind both the technology's capabilities and the concerns of the people building it. Voluntary testing frameworks and closed-door meetings with major AI labs may represent movement, but without public accountability or enforceable rules, researchers warn the gap between industry practice and genuine safety oversight could widen — especially as global competition accelerates the pace of model releases.