The U.S. government is preparing to let federal agencies test Anthropic’s Misos model as part of efforts to bolster cybersecurity. The move comes roughly 50 days after President Donald Trump ordered all federal agencies to immediately stop using Anthropic technology.
Bloomberg reported on the 16th that the administration is weighing a rollout of Misos to major federal departments.
Anthropic currently provides the model to a limited set of agencies and private firms, where it’s used to detect software vulnerabilities and strengthen cyber defenses.
Gregory Barbaccia, the White House Office of Management and Budget chief information officer, told agency chief information officers in an email that officials are considering providing agencies with a modified version of the model after implementing appropriate safeguards and protections.
No decision has been made on deployment or timing; officials say further guidance on approved uses will follow.
The review reportedly includes the Defense Department, Treasury and Homeland Security. Treasury, in particular, has pushed to adopt the model to hunt for vulnerabilities in its internal systems.
A U.S. government official warned that handing this model to a single hacker would be like turning a conventional soldier into a special-operations commando — an example of how AI can create asymmetric advantages on the cyber battlefield.
Out of concern over misuse, Anthropic has not publicly released Misos. The company has only offered the model to select agencies through a restricted program called Project Glasswing, citing its powerful code-generation and vulnerability-detection capabilities.
After a Defense Department contract fell through on Feb. 27, President Trump ordered all federal agencies to immediately stop using Anthropic technology and directed agencies already using Anthropic products, including the Pentagon, to phase them out over six months.
Policy has shifted since Misos emerged. The White House has stood up a Misos response task force, and Treasury officials have convened meetings with tech companies and major banks to coordinate defenses.
On the 15th, in an interview with Fox Business Network, President Trump said AI could weaken trust in banking systems but also acknowledged it could help make banks more secure and resilient.
Officials say the debate extends beyond a simple technology decision; it’s now a core national security issue in the AI era.
By Chan Park cpark@aitimes.com
