The Commerce Department on Friday cleared Anthropic to release its Mythos 5 model to about 100 vetted companies and federal agencies, ending a two-week standoff that had cut even the National Security Agency off from one of the most advanced U.S. artificial-intelligence systems. Hours earlier, OpenAI said it would comply with a White House request to limit its newly announced GPT-5.6 family to a small group of trusted partners before any general rollout.
The paired decisions install a case-by-case access regime as the default for frontier U.S. AI models, less than a month after President Trump signed an executive order asking developers to let the government assess capabilities before a full release. Both companies now ship their strongest systems only to organizations the administration has signed off on, with broader availability set on a timetable Washington controls.
What the letter says
Commerce Secretary Howard Lutnick laid out the Mythos 5 exception in a June 26 letter to Anthropic co-founder Tom Brown, who has taken the lead in negotiations with the administration, replacing Chief Executive Dario Amodei in that role. "I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model," Lutnick wrote. The letter did not restore access to Fable 5, the public-facing Mythos-class model that Anthropic disabled two weeks ago to comply with an export-control directive.
Anthropic spokeswoman Danielle Ghiglieri told The Verge the company had "received notice from the US government that Mythos 5, our strongest cybersecurity model, can be redeployed to a small group of cyber defenders and infrastructure providers." The directive that bars foreign nationals, including Anthropic's own foreign-national employees, from touching either model remains in force outside the approved list.
OpenAI's three-tier launch
OpenAI on Friday unveiled GPT-5.6 Sol, Terra and Luna, named for their capability tiers, and said Sol is its strongest model yet, with gains in coding, biology and cybersecurity. The company said it previewed the models and shared its rollout plans with the government before launch and is working with the administration on "a repeatable process for future model releases." OpenAI did not name the partners cleared for early access and said general availability is expected in the coming weeks.
The counterpoint
OpenAI used its own announcement to push back on the framework it had just accepted. "We don't believe this kind of government access process should become the long-term default," the company wrote. "It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them." The Verge reported that pressure on the White House had been building inside the U.S. AI industry over fears that Chinese labs could close the gap while top American models sat offline. Administration officials beyond Lutnick's letter had not publicly elaborated on the access criteria by press time.
Anthropic's broader fight is not over. The Defense Department earlier this year labeled the company a supply-chain risk, a designation historically reserved for foreign adversaries, and Anthropic's suit to reverse the blacklisting is still in litigation. Lutnick wrote that he reserves the right to adjust the license requirements if circumstances change. The next test is whether the administration approves a similar exception for Fable 5.