Trump’s Grok chats before Venezuela operation: Democrats push for public release

Lawmakers seek chatbot records

In a joint letter to Trump last week, Reps. Jamie Raskin of Maryland and Robert Garcia of California said his AI prompts and the responses should be treated as public records under the 1978 Presidential Records Act. They also questioned the use of privately operated AI systems in foreign policy.

The lawmakers alleged Trump consulted Grok, but not Congress, about military deployment to Venezuela. Public reporting cited in their letter says he spent multiple hours prompting Grok in December 2025 about how Venezuelans might respond to Nicolás Maduro’s capture, weeks before the operation and Maduro’s arrest.

Military use raises stakes

The lawmakers’ letter also raised questions about the White House’s use of chatbots in military activity abroad, including the war with Iran. It said the public should know if an AI chatbot’s faulty advice contributed to the killing of innocent schoolchildren.

Limits of AI advice

Large language models generate likely text based on training data and can produce biased or fabricated claims while sounding confident. In military settings, errors could contribute to missile strikes, arrests or civilian deaths.

AI systems can also echo users’ existing views. One journalist found earlier Grok models had “tripled” sycophancy rates compared with popular ChatGPT and Claude models, describing a change researchers at MIT identify as actively harmful.