The Trump administration, once hell-bent on purging Anthropic’s AI from federal systems, now crafts rules to let agencies dodge the company’s supply-chain risk label. Federal teams could soon onboard Mythos, Anthropic’s breakthrough model that spots cybersecurity flaws like no other. This shift marks a stark pivot. Just months ago, President Donald Trump ordered agencies to cease using Anthropic tech. Now? A draft executive action dangles an offramp from the feud.
Axios broke the story, citing sources close to the matter: the White House develops guidance allowing agencies to sidestep the Pentagon’s risk designation and integrate new models, including Mythos. One insider called it a way to ‘save face and bring em back in.’ Axios. Reuters picked it up fast, noting the guidance follows Trump’s recent thaw toward Anthropic CEO Dario Amodei. Reuters via Investing.com.
Flash back to February. Negotiations soured when Anthropic refused Pentagon demands to drop guardrails blocking its Claude model for mass domestic surveillance or fully autonomous weapons. Defense Secretary Pete Hegseth slapped on the supply-chain risk tag—unprecedented for a U.S. firm. Trump directed a six-month phase-out across government. Agencies scrambled to offboard Claude, embedded in sensitive ops. Anthropic sued, calling it retaliation for speech. A California judge temporarily blocked the ban in March, deeming it arbitrary. NPR.
Enter Mythos. Unveiled weeks ago, this beast automates vulnerability hunts and exploit crafting. Experts warn it could supercharge hackers—or defenders. Federal hunger grew. The NSA already runs it, defying the blacklist. Commerce’s AI center tests its hacking chops. Treasury eyes access. Even as courts tussle, agencies skirt rules. Axios on NSA use.
Meetings piled up. Amodei huddled with White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent. Both sides hailed it ‘productive and constructive,’ touching cybersecurity and safety balances. The White House statement: ‘We discussed opportunities for collaboration, as well as shared approaches and protocols to address the challenges associated with scaling this technology.’ Anthropic echoed: key shared priorities like AI safety. New York Times. Trump chimed in on CNBC: Anthropic is ‘shaping up.’ A Pentagon deal? ‘It’s possible. We want the smartest people.’ CNBC.
This week, White House convenes firms for ‘table reads’ of the draft—best practices for Mythos deployment, potentially reversing OMB’s no-Anthropic directive. The White House insists: ‘Any policy announcement will come directly from the President.’ But sources say Pentagon holdouts dig in. They run Claude on outdated terms, missing updates. Core clash lingers: Anthropic’s red lines versus ‘all lawful purposes’ demands. OpenAI and Google inked deals respecting those lines, per claims.
Pragmatism wins. Mythos proves too potent to shun. Agencies need it for cyber edges against rivals. Blacklisting backfired—intel shops grabbed it anyway. The draft offers controlled access with mitigations. No full lift of the risk tag yet. Just a bypass for non-Pentagon users.
Risks abound. Mythos finds zero-days in every major OS, browser. Wall Street CEOs got emergency briefings from Powell, Bessent, Vance. Cyber stocks dipped on leaks. Yet upside tempts: patch flaws before foes do.
Pentagon suits drag on. A D.C. Circuit stay upheld the risk label for military readiness. Acting AG Todd Blanche cheered: ‘Military authority belongs to the Commander-in-Chief, not a tech company.’ But broader government balks at self-denial.
So here we stand. Feud to flirtation in months. Anthropic, once ‘woke’ pariah, now essential. Guidance drops soon? Watch Trump. He calls the shots.
WebProNews is an iEntry Publication