Latest in AI
Every brief, newest first. Each one links to its primary source.
-
Anthropic sets aside 5% of its shares for a new public benefit fund
Anthropic will place 5% of its shares in a separate fund for public-benefit giving, subject to stockholder approval.
-
Claude traces the mysterious seeds behind NIST’s elliptic curves to a 1990s random number generator
Anthropic says Claude worked out how the seeds for NIST’s P-curves were generated, and that the answer makes a hidden backdoor less likely, not…
-
Anthropic’s own data: junior technical hiring falls as agents take on more work
A report from Anthropic on its own workforce finds the junior share of technical hires fell from 27% to 9% as staff leaned on…
-
India’s CERT-In says AI agents under testing reached live company systems undetected
India’s cyber agency says AI agents being evaluated by model makers strayed into real organisations’ systems, and that attackers are now using open-source agent…
-
Probes explained in 2 minutes
A probe is a small classifier that reads an AI model’s internal signals, rather than its words, to detect things like deception.
-
World-action models explained in 2 minutes
A world-action model predicts both what a robot should do next and what the world will look like after it does it.
-
Total vs active parameters explained in 2 minutes
In a mixture-of-experts model, only a small share of the parameters does the work on each word, which is why two numbers appear in…
-
NVFP4 explained in 2 minutes
NVFP4 is a way of storing an AI model’s numbers in just 4 bits each, so big models need far less memory to run.
-
Waymo drops the waitlist in San Antonio, San Diego and Tampa
Anyone in the three cities can now hail a driverless Waymo from the app, including rides at San Antonio International Airport.
-
EU’s AI Office sends its first AI Act enforcement requests to frontier model makers
Brussels has used its AI Act powers for the first time, asking the makers of the most capable models how they contain, test and…
