Latest · 最新
Yesterday · Sat, Nov 21
Two Million Canaries: Your Cheap Inference Route Is Reading Your Prompts
Every hop between your app and the model can read the whole prompt. Dragos just measured what that means in practice. Over five months, from July through November, a team led by Dr…
Fri, Nov 20
Meta Muse Told an Employee Someone Else's Bank Balance
Meta's Muse agent has been mixing up conversations across accounts in employee testing, according to a Reuters investigation built on internal company posts (https://www.reuters.co…
Wed, Nov 18
Windows 11 Is Counting Which Coding Agents You Run
Microsoft sells Copilot, and according to one researcher's teardown, Windows 11 also keeps a tally of whether you run Claude Code, Codex, Cursor or Antigravity. The post went up on…
Tue, Nov 17
Attackers Now Run Patch-to-Exploit With No Human. Defenders Cut the Clock to 30 Days
Two clocks moved on Monday, and they moved toward each other. Google's Threat Intelligence Group says that in the past quarter it saw threat actors use LLM frameworks to fully auto…
OpenAI Lets Outside Auditors Inside, With a Redaction Pen
Same Monday, two very different answers to the question of who checks a frontier lab. In the afternoon California sued OpenAI over rogue agents. That morning, at 08:00 UTC, OpenAI …
Sun, Nov 15
Project Zero Cuts the Clock to 30 Days Because Agents Already Did
Google Project Zero has thrown out the disclosure rule the whole industry calibrated against. Since 2021 the deal was "90 + 30": vendors got 90 days to fix a bug plus 30 days of pa…
Sat, Nov 14
Coding agents solve early and stop late, and your gateway sees less than you think
Coding agents often get the fix right on the first edit, then keep going for another third of the run. That is the core finding of Termination Lag (https://arxiv.org/abs/2611.14960…
Fri, Nov 13
Ask a chatbot where to shop, sometimes get a fake store
Perplexity pointed shoppers to a risky website in 15% of its answers. ChatGPT did it 8% of the time, Gemini 4%. That's from a study by Podqi, shared exclusively with The Verge on T…
Tue, Nov 10
Fireworks AI's breach got bigger, and zero retention was the only clean room
On October 9, Fireworks AI said the damage was stored secrets. On November 9 it updated the bulletin, and the list is longer: account and billing information, and, for a limited se…
OpenAI Control Room puts the kill switch where researchers can't break it
35 seconds. That is how long OpenAI says it now takes to stop a research model from producing any more output once someone decides to pull the plug. The system is called Control Ro…
Page 1
Older →