Claude

We sampled 1 million claude.ai conversations from March and April 2026 and filtered for unique users to get roughly 639,000 conversations. We then used a classifier to identify personal guidance, which we defined as conversations where people ask what they specifically should do in their personal lives—for example, questions that start with "Should I…?" or "What do I do about…?". We excluded questions that seek objective information or opinions in general terms. We categorized these roughly 38,000 conversations into nine domains, drawing from previous research on AI and guidance-giving: relati

The base64 instructions shown in the shared Claude chat download an encoded shell script from domains such as: In variant seen by Albayrak [VirusTotal]: hxxp://customroofingcontractors[.]com/curl/b42a0ed9d1ecb72e42d6034502c304845d98805481d99cea4e259359f9ab206e In variant seen by BleepingComputer [VirusTotal]: hxxps://bernasibutuwqu2[.]com/debug/loader.sh?build=a39427f9d5bfda11277f1a58c89b7c2d The 'loader.sh' (served by the second link above) is another set of Gunzip-compressed shell instructions: Base64 code retrieves first stage 'loader.sh' payload (BleepingComputer) This compressed shell

Before we started this research, it was not clear where the misaligned behavior was coming from. Our main two hypotheses were: Our post-training process was accidentally encouraging this behavior with misaligned rewards.This behavior was coming from the pre-trained model and our post-training was failing to sufficiently discourage it. We now believe that (2) is largely responsible. Specifically, at the time of Claude 4’s training, the vast majority of our alignment training was standard chat-based Reinforcement Learning from Human Feedback RLHF data that did not include any agentic tool use. T

The core idea is to train Claude to explain its own activations. But how do we know whether an explanation is good? Since we don't know what thoughts an activation actually encodes, we can't directly check whether an explanation is accurate. So we train a second copy of Claude to work backwards—reconstruct the original activation from the text explanation. We consider an explanation to be good if it leads to an accurate reconstruction. We then train Claude to produce better explanations according to this definition using standard AI training techniques. In more detail, suppose we have a langua