HPSCI Ranking Member Calls Intelligence Agency Access to Frontier AI Models a National Security Imperative
Rep. Jim Himes (D-Conn.), HPSCI ranking member, told a Politico Security Summit panel on May 12 that it would be "insane" for NSA and the IC to lack early access to the most capable AI models for hacking and cyberdefense. Multiple unnamed sources told Nextgov/FCW and Government Executive that NSA has been testing
The NSA's confirmed testing of Mythos, an Anthropic model held from public release for its substantial cyber capabilities, establishes that the IC has secured meaningful early access to frontier AI notwithstanding the ongoing DOD-Anthropic dispute. That dispute's trajectory, not its current state, is the operative risk: present contracting friction becomes a national security liability only if Hegseth escalates beyond commercial logic into institutional retaliation. Draft policy language circulating within the administration, aimed at clarifying government authority to use commercial AI without vendor-imposed stipulations, signals an internal effort to foreclose that path. The IC is unlikely to face material disruption to frontier model access through the end of 2026, given NSA's active testing foothold and bureaucratic momentum toward a contracting fix. The scenario that produces disruption requires an escalatory human choice with no current firm indicator.
4 sources
- It would be insane for spy agencies to lack early access to AI models lawmaker says -
Defense One - 'It would be insane' for spy agencies to not have AI model early access, lawmaker says -
Nextgov/FCW - 'It would be insane' for spy agencies to not have AI model early access, lawmaker says -
Government Executive - Full Interview with Rep. Jim Himes (D-Conn.) | POLITICO Security Summit