The Trump administration has postponed an executive order requiring pre-release government security reviews of artificial intelligence models, with the president citing concerns that specific language in the draft could hinder American AI leadership, according to TechCrunch AI.
The decision marks a significant departure from stricter AI governance approaches and introduces fresh uncertainty into the regulatory landscape as major technology firms prepare to deploy increasingly capable foundation models throughout 2026.
According to the reporting, Trump specifically objected to language in the draft order, stating he did not want to “get in the way of that leading” — an apparent reference to maintaining US dominance in AI development. The administration has not provided a revised timeline for the order’s implementation.
The delayed executive order would have required AI developers to submit models for government security assessments before public release, a measure intended to identify potential national security risks, including capabilities that could be exploited for cyberattacks, biological weapons development, or other malicious purposes.
Regulatory vacuum emerges
The postponement creates a governance gap at a critical juncture. OpenAI, Anthropic, Google DeepMind, and other leading labs are preparing to release models with significantly enhanced capabilities over the coming months. Without mandatory pre-release reviews, these systems will reach market subject only to voluntary company commitments and existing sectoral regulations.
The decision follows the Trump administration’s January 2025 revocation of President Biden’s October 2023 executive order on AI safety, which had established reporting requirements for developers of large-scale models. That order mandated that companies training models using more than 10^26 floating-point operations notify the government and share safety test results.
Industry groups had lobbied against prescriptive pre-release review requirements, arguing they would slow innovation and place American firms at a competitive disadvantage relative to Chinese developers operating under different constraints.
Business impact
The delay represents a clear win for major AI developers, who gain operational flexibility and faster time-to-market for new models. Firms including OpenAI, Anthropic, and Google can now proceed with product launches on their own timelines, subject only to voluntary safety commitments.
Conversely, AI safety organisations and national security advocates lose a potential mechanism for systematic risk assessment. The absence of mandatory government review shifts responsibility for identifying dangerous capabilities entirely to company internal processes, which vary significantly in rigour and transparency.
For enterprise customers, the regulatory uncertainty complicates risk management. Without government-validated security assessments, organisations deploying frontier AI models must conduct their own due diligence or rely on vendor assurances.
Capital markets have responded positively to signals of lighter-touch AI regulation, with public AI-focused equities gaining ground on expectations of reduced compliance costs and faster commercialisation cycles.
International implications
The US approach now diverges sharply from the European Union’s AI Act, which entered into force in August 2024 and imposes strict requirements on high-risk AI systems, including conformity assessments before market placement. This regulatory fragmentation creates compliance complexity for firms operating across jurisdictions.
China’s approach combines state oversight of large models with aggressive support for domestic champions, whilst the UK has maintained a sector-specific, principles-based framework. The absence of US federal requirements may accelerate calls for international coordination mechanisms, though prospects for binding agreements remain limited.
What comes next
The administration’s next moves on AI governance will be closely scrutinised. Industry observers will watch for whether revised executive order language emerges that addresses Trump’s stated concerns whilst preserving some oversight mechanism, or whether the administration opts for a purely voluntary framework.
Congressional action remains possible, with bipartisan interest in AI safety legislation, though deep divisions over the appropriate regulatory approach make comprehensive federal legislation unlikely in the near term.
The delay underscores the tension between maintaining technological leadership and managing emerging risks — a balance that will define AI governance debates as model capabilities continue their rapid advance.
