Anthropic expands Mythos access to 150 new orgs, up from the original 50
www.anthropic.com/news/expandi...
Several secretaries of state tell Bloomberg that intelligence agencies are no longer briefing them on election security threats, with one also saying that CISA has stopped conducting some security tests for states. www.bloomberg.com/news/newslet...
Catalin Cimpanu
Eric Geller
Classified briefings and some cybersecurity services have ceased.
Anthropic earlier week have a further 100 groups in 15 countries access to Mythos. My colleague Mickey Djuric confirmed that Canada, another Five Eyes member, has access to the powerful tool: subscriber.politicopro.com/article/2026...
Details: Establishes process for frontier companies to voluntarily submit AI models for federal testing 30 days before release; establishes federal AI cyber clearinghouse to track vulnerabilities discovered by AI models; and tasks CISA, NCD, OMB with finding federal AI funds.
Russian hackers are shifting their tactics from destructive cyberattacks to quieter espionage schemes aimed at Ukrainian defense efforts, as the war drags into a fifth year. More at the top of today’s NatSec Daily: www.politico.com/newsletters/...
In today’s Morning Cyber: New Zealand’s cyber agency is among the international gov agencies recently given access to Anthropic’s Mythos AI model, and Australia is also testing out the tool. Both countries are members of the Five Eyes intel sharing alliance.
Anthropic released “Mythos-class” Fable 5 for general use today, and separately rolled out Mythos 5 for a select group of agencies and cyber firms. Anthropic said Mythos 5 has “the strongest cybersecurity capabilities of any model in the world.” More w/ Dana Nickel: www.politico.com/news/2026/06...
A draft of a major bipartisan AI bill, set to be unveiled later today, includes a clause that would renew the CISA 2015 law - crucial for private sector companies sharing threat data with Feds - through 2035. Read more from my colleagues here: www.politico.com/news/2026/06...
New: Democrats on the House Homeland Security Committee are pressing DHS Secretary Mullin for details on CISA’s election security efforts - paused last year - with the midterms fast approaching: subscriber.politicopro.com/article/2026...
ICYMI over the long weekend: Researchers with early access to Anthropic’s Mythos and OpenAI’s GPT-5.5 say these models’ hacking abilities are “more” than expected — and warned that defenders are running out of time to secure networks. With my colleague Aaron Mak: www.politico.com/news/2026/05...
Researchers who have tested Anthropic’s Mythos and OpenAI’s GPT-5.5 say their hacking capabilities are a “game-changer.”
The draft legislation represents Republicans’ last realistic chance to craft federal rules governing artificial intelligence before the midterm elections.
Congress gave CISA almost $40 million in recently enacted appropriations to secure upcoming elections, but it is not yet clear how these funds will be used.
