Simon Willison·· 7 h ago
How We Took Over ChatGPT Accounts
How We Took Over ChatGPT Accounts
AI summary
How We Took Over ChatGPT Accounts Parameter Labs describe an exploit they found against ChatGPT's implementation of the artifacts pattern (OpenAI used to call this "Canvas" but it looks like the latest revision is called HTML previews ), where interactive custom HTML apps are displayed in a sandboxed iframe.
Selection record
Threshold 76Media and individualsFirst 62Second 68
Not admittedSum of both 130 < twice the threshold 152
- Source tier
- Media and individuals; this tier's threshold is 76
- Pre-filter
- passed：ChatGPT账户漏洞利用与AI安全
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: Simon Willison · simonwillison.net