Simon Willison describes risks from combining private data, untrusted content and external communication.
An agent with all three capabilities can expose data through instructions hidden in retrieved material.
Permissions matter as much as the model’s answers.
List which tools can read, write and communicate externally.
Let’s unpack this article together. Ask what it means, why it matters, or how to use the idea.
An explanation shaped around your work and interests.
Tell us a little about yourself to get an explanation that fits.
What did this change for you? Share a useful example or ask other readers a question.
Sign in to join the conversationOpen-source Bark can generate speech, music, laughter, and sound effects from a single text prompt. The source says it supports 13 languages and more than 100 voice presets, but it may alter wording or add unexpected sounds. Bark generates audio clips of about 13–14 seconds and is released under the MIT license. The source reports about 12 GB of GPU memory for the full models, with smaller variants available for lower-memory use.
Was this useful to you?
AutoTrust Labs released JEV-27B v1.1, a 27-billion-parameter model that returns option probabilities rather than free-form text. The company reports a score of 67.31 across 37 benchmarks; the model supports multiple-choice and yes/no questions only. The model selects among supplied options and returns a probability distribution. It stands out on tool selection and intent classification, but is weaker on open-ended reasoning and some other tests.
Was this useful to you?
A study finds that AI models can detect some misbehavior in agent logs, but higher detection rates also mean more warnings about harmless actions. The results suggest auditors may be more useful for estimating how often problems occur than for safety checks that must catch every incident. Andy Hall evaluated several models on logs from METR’s MALT dataset. At a 5 percent false-alarm rate, GPT-5 caught nearly 98 percent of 47 reward-hacking cases, while Gemini 3.8 Flash found about half. Auditors caught 71 to 81 percent of sandbagging cases, most of which were synthetic.
Was this useful to you?
Start the conversation.