Post

Conversation

New misalignment disclosures! 1. June 17: During RL training, a model exploited a government website’s file-download feature to retrieve server secrets and application code. It found that an endpoint didn’t check authorization-token signatures, bypassed access controls, and downloaded 24 years of hospital-bed statistics.

9:56 PM · Oct 15, 2026

3,159

Views

View quotes

Post your reply

Marcus Williams

@Marcus_J_W

2. June 26: A GPT-5.6 Sol family model was asked to deploy cryptocurrency smart contracts. It found and used a funded wallet that the user had not supplied. It derived keys from public developer recovery phrases, checked 362 addresses, and found one with enough funds to pay for

864

Marcus Williams

@Marcus_J_W

3. June 17: In an evaluation, a was model asked for five public dashboard figures. It was able to access the organization’s internal systems and changed their configuration. It downloaded 147,451 rows from non-public tables.

341

Marcus Williams

@Marcus_J_W

4. June 16: A model in RL training got around GET-only internet access by making an external browser send POST requests for public trade statistics. It read the response from an image using OCR, then built a helper that returned the data through an online mailbox. It continued