Research we co-authored on subliminal learning—how LLMs can pass on traits like preferences or misalignment through hidden signals in data—was published today in @Nature
. Read the paper: https://
nature.com/articles/s4158
6-026-10319-8
…
SAFETY
-
LLMs Pass Hidden Traits Through Subliminal Learning Signals
By
–
-
Gemini 3.1 Flash TTS Audio Now Watermarked with SynthID
By
–
All audio generated by Gemini 3.1 Flash TTS is watermarked with SynthID!
-
AI Security Framework: Sandbox, Allow-lists, Access Control
By
–
That was the case in December. 4 months and thousands of work hours later, we have a great security concept; you can go all yolo, use a sandbox (Docker or OpenShell), there are allow-lists and per-access exec allow/deny prompts. There’s hundreds of security researchers that
-
The Model Preferred Aesthetics Over Actual Function
By
–
2/5 Turns out the model wasn't remembering the solution, but it was identifying "gold-like" aesthetics like minimality & clarity. Total form over function kind of scenario.
-
LLM Judge Rejects Functional Fix for Code Aesthetics
By
–
3/5 An example: In instance psf__requests-1724, the gold fix is 2 lines. Our agent’s functional fix was 8 lines. The LLM judge rejected the correct 8-liner as "messy" and "redundant," choosing a clean but **non-functional** fix instead. See full patch in the blog:
-
LLM judges reject functional agent fixes as messy code
By
–
3/5 An example: In instance psf__requests-1724, the gold fix is 2 lines. Our agent’s functional fix was 8 lines. The LLM judge rejected the correct 8-liner as "messy" and "redundant," choosing a clean but **non-functional** fix instead. See full patch in the blog:
-
Model Aesthetics Over Function Recognition Bias
By
–
2/5 Turns out the model wasn't remembering the solution, but it was identifying "gold-like" aesthetics like minimality & clarity. Total form over function kind of scenario.
-
Interesting AI Safety Approach Gains Attention
By
–
Definitely one of the more interesting approaches to AI safety I've seen recently
-
USB Port Security Incident on HMI Device Highlights Default Settings Risk
By
–
The phone charging incident had no malicious intent. The operator needed power, there was an open USB port on an HMI, and the device's default tethering setting did the rest.
-
Physion-Eval: Benchmarking Physical Realism in AI-Generated Videos
By
–
How physically realistic are our AI-generated videos? Physion Labs, Stanford University, MIT, Harvard University, and Character AI introduce Physion-Eval. This new benchmark uses expert human reasoning to meticulously diagnose and explain physical realism failures in