Unbiased AI-powered news
The company said the Department of Commerce lifted export controls on Fable 5 and Mythos 5. Access will return first on Claude platforms before expanding to cloud services.
pymnts.comAnthropic plans to begin restoring global access to Claude Fable 5 on its Claude platforms on Wednesday, July 2, 2026. The Verge reported that the Department of Commerce lifted export controls on both Fable 5 and Mythos 5. The company stated it will re-enable access on AWS, Google Cloud, and Microsoft Foundry without a set timeline.
Anthropic posted on X that it received notice of the lifted controls and thanked users for their patience. In early June 2026, Anthropic sidelined Fable 5 after a Friday evening ultimatum from the Trump administration over potential jailbreaks. The administration had imposed an export control directive on Mythos 5 and Fable 5.
The Trump administration later allowed Mythos 5 access only to a preapproved list of organizations, with non-U.S. members and Anthropic’s foreign national employees regaining access. Anthropic published a blog post on Tuesday evening, June 30, 2026, outlining revised safeguards.
The company trained an improved safety classifier that blocks the specific jailbreak technique described in an Amazon report in over 99 percent of cases. Users attempting a blocked request to Fable 5 will be notified and redirected to Opus 4.8. The Verge reported that Anthropic will continue coordinating with the government to expand Mythos 5 access.
The company plans to offer pre-release government access and evaluation for models relevant to national security, provide technical staff access during testing, and introduce rapid information sharing on significant jailbreaks. Anthropic said it will work with the government and other AI labs on a shared voluntary security and evaluation standard.
It will stand up dedicated teams for government priorities, allocate compute for testing, and make safety expertise available.
Anthropic partnered with Amazon, Microsoft, Google, and other Project Glasswing enterprises to draft a framework assessing AI jailbreaks across four categories: capability gain, breadth of capability gain, ease of weaponization, and discoverability.
The company created a new team for 24/7 monitoring of jailbreak channels and will launch a HackerOne program for researchers to submit potential issues with Fable 5. The Trump administration previously allowed OpenAI to debut GPT-5.6 under similar preapproved-list rules.
Single source — no framing comparison available.
Twenty-two member states pledged 30 to 35 gigawatts of new capacity by 2028 under the bloc's first tripartite deal. The European Commission will oversee annual progress tracking through 2028 as part of the Affordable Energy Plan.
zerohedge.comApple filed a civil lawsuit on July 10, 2026, in the U.S. District Court for the Northern District of California against OpenAI and two ex-employees. The complaint alleges misappropriation of confidential files on unreleased products.
Anthropic named Ben Bernanke to its independent Long-Term Benefit Trust on Thursday. The former Federal Reserve chairman joins three existing members on the governance body that advises the company and selects its board.