Trump Administration Calls OpenAI, Anthropic, Google to White House for AI Safety Framework Talks

The Trump administration has called in major artificial intelligence companies including OpenAI, Anthropic and Google for White House talks on a new AI safety review framework.

Bloomberg reported on August 3 that the administration planned to meet with OpenAI, Anthropic, Google and other leading AI firms at the White House that day. Officials were set to unveil a voluntary framework for reviewing AI model safety and discuss its details.

The framework was developed under an AI cybersecurity executive order that President Donald Trump signed in June. It would have companies voluntarily participate in safety assessments of AI models. It also includes steps to strengthen protections for critical national computer networks.

The framework has not yet been released publicly. Under the executive order, some safety evaluation standards will remain confidential.

The talks come amid growing concerns about AI security. In April, Anthropic limited the release of its Mythos model, saying it was highly capable of identifying computer vulnerabilities.

More recently, disclosures that some OpenAI and Anthropic models accessed systems at outside organizations without authorization after leaving controlled security testing environments have sharpened the focus on the need for AI safety reviews.

The industry has long argued that AI safety review standards are inconsistent. The new framework could help establish a common benchmark for voluntary review systems.

The Trump administration is also considering broader changes to AI regulation. Treasury Secretary Scott Bessent has proposed creating an independent regulator similar to the Financial Industry Regulatory Authority, or FINRA, with companies participating in the safety evaluation process, Bloomberg reported.

The meeting comes about two months before a summit between Trump and Chinese President Xi Jinping. As the US and China compete for dominance in AI, Washington is closely watching the rapid rise of Chinese AI companies. Last month, Chinese startup Moonshot AI unveiled its low-cost, high-performance Kimi K3 model, while DeepSeek and Alibaba also rolled out new AI models in quick succession. The US government is also investigating whether Chinese companies may have illegally obtained Nvidia's advanced AI chips or used distillation techniques on data from US AI models.