White House to Add Open AI Models to Safety Testing Framework
The Trump White House is expected to broaden its voluntary AI safety-testing framework in the coming weeks, bringing open models under the same pre-release review currently applied to frontier closed systems from labs like Anthropic and OpenAI, according to WIRED via an X post. Open models would be added once they reach the 'frontier' capability level of those top proprietary systems. This shift extends government oversight to publicly available models, a category previously outside the framework. However, the framework remains voluntary, and the exact trigger for inclusion is undefined, leaving adoption by open-source developers uncertain. The update signals increasing regulatory attention on open-weight AI as its capabilities approach proprietary benchmarks.
White House to Add Open AI Models to Safety Testing Framework
Trump White House is expected to expand its AI oversight framework in the coming weeks and months — including adding open models to the framework as soon as they reach the same “frontier” capabilities of Anthropic and OpenAI’s top models. WIRED says the voluntary framework currently covers frontier closed models from labs such as OpenAI and Anthropic.
Key takeaway
Open-source AI models will soon face voluntary safety testing if they match frontier capabilities, marking a notable expansion of White House oversight beyond closed labs.
What happened
According to an X post from LLMgram-owned news, the Trump White House is expected to expand its AI oversight framework to include open models. The post attributes the reporting to WIRED, stating that open models will be added once they achieve the same 'frontier' capabilities as Anthropic's and OpenAI's leading models.
The current voluntary framework already covers frontier closed models from labs such as OpenAI and Anthropic, as per WIRED. The expansion would bring open models under the same prerelease safety-testing process, though the exact timeline and criteria remain unspecified.
Evidence
White House expected to expand AI oversight framework to include open models once they reach frontier capabilities
X · attributed
Trump White House is expected to expand its AI oversight framework in the coming weeks and months — including adding open models to the framework as soon as they reach the same “frontier” capabilities of Anthropic and OpenAI’s top models.
Current framework voluntarily covers frontier closed models from labs like OpenAI and Anthropic
X · attributed
WIRED says the voluntary framework currently covers frontier closed models from labs such as OpenAI and Anthropic.
Why it matters
This move could set a precedent for how governments treat open-weight models, potentially affecting deployment, liability, and research in the AI community.
Limits and uncertainties
The reporting is based solely on an X post citing WIRED; the original WIRED article is not provided, so details may be incomplete or paraphrased.
The framework is voluntary, so compliance by open-model developers is not guaranteed, and the exact criteria for 'frontier' capability are unspecified.
Practical implications
Open-model developers may need to plan for potential safety evaluations if their models approach frontier performance, and might consider proactive engagement with the framework.
Organizations relying on open models should monitor policy changes to assess compliance risks and adjust deployment strategies accordingly.
What to watch
Look for official White House announcements or WIRED's full report for concrete criteria and timeline regarding when open models will be included.
Monitor whether open-source labs like Meta or Mistral volunteer for testing under the expanded framework.
