The United States Department of Defence has reportedly told officials that Anthropic artificial intelligence tools may still be used beyond a planned six-month phase-out period in limited cases.
According to an internal memo, the Pentagon AI policy will allow exemptions when the technology is essential for mission-critical national security operations. The memo, signed by Department of Defence Chief Information Officer Kirsten Davies, outlines how certain teams may continue using Anthropic tools if no viable alternative exists.
The memo says that exemptions will be given only in “rare and extraordinary circumstances”. A risk-mitigation plan from Pentagon offices that need approval will be required, which will explain why Anthropic’s AI technology is needed for national security work.
The announcement was made weeks after the Pentagon made a decision to bar the company’s technology from its systems and contractor networks. The decision was made due to a disagreement over the guidelines for the use of artificial intelligence by the military in defence operations. The company was considered a potential risk in the supply chain.
Experts believe that the memo highlights how hard it may be to fully remove the company’s technology from government systems. According to Reuters, a government contracts lawyer with McCarter and English Franklin Turner said that contractors may have trouble ensuring that their software does not contain open-source code associated with Anthropic's tools.
Turner further added that many organisations that work with the Pentagon may be able to obtain waivers to use the technology until alternatives are developed.
The matter has now reached the courtroom as well. Anthropic has filed a lawsuit to prevent the banning of its technology by the Pentagon.