White House releases watered-down executive order on AI

The White House released a weakened executive order on AI Tuesday, reducing the voluntary government review window for new models from 90 days to 30 days after intense lobbying by tech executives convinced President Trump to kill a stronger version. The order, which replaces a previously reported proposal that would have given the government access to unreleased frontier models for safety checks, instead creates an "AI cybersecurity clearinghouse" and directs national security agencies to develop a classified benchmarking process for new models.

White House releases watered-down executive order on AI The White House released a weakened executive order on AI on Tuesday, a little more than a week after killing a previous version of the order after what was reportedly intense, direct lobbying https://sherwood.news/power/report-white-house-ai-oversight-executive-order-doa/ of the Oval Office by tech executives. The order’s most significant change to what was reported in late May is a shortened window of voluntary government review of new models from 90 days to 30 days. After Anthropic’s Mythos model spooked companies https://sherwood.news/tech/anthropic-our-new-mythos-model-is-so-powerful-we-cant-release-it/ and governments around the world, the White House was reportedly ready to respond with an executive order that would have given the government access to unreleased frontier models for up to 90 days https://sherwood.news/power/report-white-house-informed-ai-companies-about-plans-for-government-to-vet-new-models/ before public release, to ensure safety. Top AI companies were briefed https://sherwood.news/power/report-white-house-informed-ai-companies-about-plans-for-government-to-vet-new-models/ on the proposed executive order, and a White House event with an extensive roster of tech executives was ready to go, but it was killed at the last minute, according to reports. Axios reported that last-minute lobbying by former White House AI and Crypto Czar David Sacks, along with other tech executives, helped convince President Trump to kill the order https://sherwood.news/power/report-white-house-ai-oversight-executive-order-doa/ . Trump told reporters, “I didn’t like certain aspects of it. I postponed it.” The now finalized order calls for the creation of an “AI cybersecurity clearinghouse” in concert with the AI industry, and directs national security agencies to develop and maintain a “classified benchmarking process” to review the capabilities of new frontier models. After Anthropic’s Mythos model spooked companies https://sherwood.news/tech/anthropic-our-new-mythos-model-is-so-powerful-we-cant-release-it/ and governments around the world, the White House was reportedly ready to respond with an executive order that would have given the government access to unreleased frontier models for up to 90 days https://sherwood.news/power/report-white-house-informed-ai-companies-about-plans-for-government-to-vet-new-models/ before public release, to ensure safety. Top AI companies were briefed https://sherwood.news/power/report-white-house-informed-ai-companies-about-plans-for-government-to-vet-new-models/ on the proposed executive order, and a White House event with an extensive roster of tech executives was ready to go, but it was killed at the last minute, according to reports. Axios reported that last-minute lobbying by former White House AI and Crypto Czar David Sacks, along with other tech executives, helped convince President Trump to kill the order https://sherwood.news/power/report-white-house-ai-oversight-executive-order-doa/ . Trump told reporters, “I didn’t like certain aspects of it. I postponed it.” The now finalized order calls for the creation of an “AI cybersecurity clearinghouse” in concert with the AI industry, and directs national security agencies to develop and maintain a “classified benchmarking process” to review the capabilities of new frontier models.