OpenAI CEO Sam Altman is expected to return to Washington on Wednesday and Thursday for meetings with the White House and other administration officials as well as bipartisan members of Congress to discuss artificial intelligence policy, according to three people close to the industry and the administration, granted anonymity to detail the meetings.
Altman’s visit comes at a high-stakes moment not only for OpenAI but any AI company with frontier models, as lawmakers and regulators grapple with balancing the need to ensure the models can’t be used to do harm while at the same time allowing innovation to flourish. And over it all hangs concerns about threats from China.
Altman is expected to meet with Treasury Secretary Scott Bessent, one of the people said, and Commerce Secretary Howard Lutnick, two of the people said. He’s also expected to meet with Sen. Mark Warner (D-Va.), the top Democrat on the Senate Intelligence Committee and a co-chair of the Senate’s Cybersecurity Caucus, POLITICO reported earlier Monday.
In his meetings on both ends of Pennsylvania Avenue, Altman will preview OpenAI’s newest models, the company said last week. He’s also likely to face questions over a recent cybersecurity attack conducted by OpenAI’s latest models without human direction — the first such documented incident of its kind. Altman will also discuss the implementation of the White House’s June executive order stipulating heightened government scrutiny of advanced models.
The June executive order’s 60-day period to develop the framework will draw to a close on August 1. It calls on companies to voluntarily submit advanced AI models to the government 30 days before release for safety testing, and then release the models to government-approved partners for initial limited rollouts before releasing them to the wider public.
The deadline has set off an intense lobbying effort by the country’s largest AI companies, which have resumed discussions with the White House over the past two weeks after export controls on Anthropic and restrictions on OpenAI’s latest model were lifted, the three people familiar with the discussions said.
OpenAI, Anthropic and Google jointly submitted extensive “redline” edits to a draft framework from the White House's Office of the National Cyber Director earlier last week, according to one of the people, who is close to the administration.
"They produced one joint red-line document," the person said, describing it as an unusual display of unity among companies.
The three companies declined to comment.
One proposal backed by at least some industry players would require review only for the very highest-performing AI systems available industry-wide, meaning some companies’ best models would not undergo a review at all. The executive order will determine the exact benchmark performance that qualifies a model for government review.
Companies have been hungry to return the focus to the June 2 executive order in hopes that it will bring stability to the industry, which has been rattled by recent abrupt AI policy decisions.
Shortly after the president signed the June order — but before it was implemented — the administration slapped export controls on Anthropic, citing safety concerns. The company disputed the administration’s assessment but ultimately removed its Fable and Mythos models from the market for two weeks while it negotiated a path to lifting the export controls.
The administration also stepped in to request OpenAI amend the rollout of its ChatGPT-5.6 model to only government-approved partners, which Altman at the time said was not the company’s “preferred” way of releasing the model and ideally would be alleviated by implementing a more “sustainable” government process to guide companies.
from White House https://ift.tt/mzfu1aL
https://ift.tt/hut3JYN
0 Comments