"Anthropic is releasing a general-use Mythos-class AI model with better capabilities than anything it's published before, Axios' Madison Mills and Sam Sabin report.

The company once deemed this new class of model so disruptive that existing safeguards were insufficient, and it restricted access to just a handful of trusted organizations.
Now it's releasing a public version less than three months later.

The new Fable 5 model is better at knowledge work, software engineering, scientific research and more outscoring competing models from OpenAI and Google DeepMind, Anthropic says.

It includes safeguards to prevent coders from using it to hack infrastructure or ask about sensitive biological capabilities.

Dianne Penn, Anthropic's head of product management, research and labs, says the company ran Fable 5 through internal and external testing to ensure safety.
Axios

huh. tnx B but, no, ai isn't for me.

"Data centers now seek amounts of electricity that used to be associated with entire cities raising questions about who pays for new infrastructure, who gets access to scarce power and how quickly projects can connect to the grid."
Axios

"It was a historic, even jarring, scene that captures a once-unimaginable geopolitical ordering. The world's most powerful heads of state gathered in the French Alps this week for the annual G7 summit, with the CEOs of America's dominant AI companies seated and treated like heads of nation-states themselves.

Why it matters: This is the future many leaders and AI CEOs envision heads of state and the masters of tech in constant discussion, and sometimes conflict, over who controls AI, its rules, and its application to governing and world security.

Think of Anthropic vs. Trump as merely a small test run of this dynamic, with governments battling private companies over their products' threat to U.S. or global security.

AI CEOs sat around the table with leaders of the world's democracies, treated as peers. The companies, creating the world's future economy and security infrastructure, are now the equivalent of nation-states."
Axios


Sam Altman, seen here declaring France deleted from the grid

AI-powered phishing kits that automate hackers' workflows are fueling a huge increase in advanced cyberattacks, Sam Sabin writes from a new report shared first with Axios.

The report, from security firm Huntress, finds a nearly 1,400% rise in "device-code" phishing attacks in the first four months of 2026 compared to the second half of 2025.

In such an attack, victims are directed to a real Microsoft login page and asked to enter a code sent by the attacker.

Once the victim logs in and verifies themselves with that code, the attacker gets a token granting them access.

Huntress says that the tools needed for such an attack are now readily available in AI-powered kits sold to hackers as a subscription package.

Those toolkits let attackers with "little to no technical skill" launch sophisticated phishing campaigns, the report finds.

Another finding: No two phishing messages were identical across hundreds of incidents analyzed, suggesting cybercriminals are using generative AI to personalize "lures" sent to victims at massive scale."
Axios


Pictured here, a criminal phishing

Nvidia
Micron
Broadcom
SanDisk
SK Hynix
Samsung
AMD
Intel