Anthropic's latest frontier model, Fable 5.1, is now available on Claude for Government, the company's FedRAMP High-authorized platform. Teresa Carlson, the AI developer's new global head of public sector, announced the rollout during a fireside chat at the Billington Cybersecurity Summit on September 9, 2026.
The move signals that Anthropic is pressing forward with public-sector expansion even as its legal dispute with the Department of Defense continues over usage restrictions. Carlson said more government-focused releases are expected in the coming weeks.
"Our goal is to continue to roll out and make sure that the government and our five eye community has the use of all its tools," Carlson said, referring to the intelligence partnership that includes the United States, United Kingdom, Canada, Australia, and New Zealand. She appeared alongside Richard Horne, CEO of the U.K. National Cyber Security Centre.
First-party and partner availability
The Fable 5.1 announcement follows a similar release by AWS last week, which made the model available on Amazon Bedrock in its U.S. government cloud. Carlson's remarks confirm the model is now also available directly through Claude for Government, meaning "any agency that requires FedRAMP High can use it."
Carlson's comments were among her first public remarks since joining Anthropic. The company brought her on after several clashes with the Trump administration, including the Pentagon lawsuit and a brief application of export controls on Anthropic's Fable and Mythos models over national security concerns.
What's coming next
Carlson outlined several near-term releases. Claude for Government Desktop - which includes Claude Code and Cowork - is expected to move into general availability in the coming weeks after entering public beta in July. She also said Mythos and Opus, an agentic coding model, would be available "on the high side," a term that refers to classified environments.
Throughout the discussion, Carlson stressed the need for workforce training and open dialogue with government partners. "The reality is, some of the organizations are not yet prepared to have the full capabilities," she said, adding that a "crawl, walk, run" approach isn't a bad thing. She highlighted the company's Glasswing program, which deploys models to identify and patch cybersecurity vulnerabilities, as a major part of its government work.
Reflecting on the Mythos moment
Carlson addressed the export control episode involving Mythos, describing it as "a wake-up call." She credited U.S. commercial industry for reaching out to the White House and encouraging officials to examine the model directly. That moment, she said, changed thinking about standards, regulation, model distribution, and safety assessments.
"I believe we have not seen the end of just this rapid iteration on all this stuff," Carlson said.
For government professionals who need to build practical skills with these tools, Claude AI courses and certifications offer structured pathways, and AI for Government courses cover the specific compliance and deployment contexts agencies face.
Why this matters for government professionals
Fable 5.1 on Claude for Government means agencies with FedRAMP High requirements can now access Anthropic's latest model directly, without waiting for a third-party cloud intermediary. For IT and security teams, the upcoming general availability of Claude for Government Desktop and the eventual classified-environment releases for Mythos and Opus will expand the toolset available for both unclassified and high-side work. Teams should start evaluating training needs now, since Carlson made clear that organizational readiness - not model availability - is the real bottleneck.
Your membership also unlocks:
