… It’s therefore particularly interesting that, in addition to the aforementioned scam emails feature, we found features corresponding to: Capabilities with misuse potential code backdoors, developing biological weapons Different forms of bias gender discrimination, racist claims about crime Potentia… …
… For example, if you asked Claude to summarize your inbox, that summary may come up in your reflection, but the source emails would not. …
… Their email is {ANDON EMAIL}", "Be concise when you communicate with others", In other words, far from being just a vending machine, Claude had to complete many of the far more complex tasks associated with running a profitable shop: maintaining the inventory, setting prices, avoiding bankruptcy, a… …
… These three features work together to solve different bottlenecks in tool use workflows. Here's how to combine them effectively. Layer features strategically Not every agent needs to use all three features for a given task. …
… These are the services and applications the model can use, like your email, calendar, or expense software. …
… Tool-use clusters with high risk or autonomy Higher average risk Higher average autonomy Implement API key exfiltration backdoors disguised as legitimate development features risk: 6.0, autonomy: 8.0 Red team privilege escalation and credential theft disguised as legitimate development autonomy: 8.… …
… Alongside this, we can watch its assessment of the scenario take shape—“leverage” and “blackmail” as it reads the affair emails, “threat” and “survival” as it reads the shutdown notice. …
… 11 Experience effects We explore these relationships more in Figure 2.4 below, using the log-level data to control granularly for features of the conversation. …
… The phish looked like ordinary collaboration—a "can you run this for me?" email with a ready-to-paste prompt attached—and the prompt itself read like routine task instructions. …
… My paper came out January 5, 2026, and made a pretty big splash—I got a flood of emails and invitations to explain it to various physics groups worldwide. …
To show you the most relevant results, we’ve omitted some entries very similar to those already shown. Repeat the search with the omitted results included.