ChatGPT web search leaked hidden debug payloads via SSE streams, OpenAI has patched it
metehan777 · x · 2026-11-07
- A talk revealed that every ChatGPT web-search answer carried a hidden debug payload with queries, scores, prompt blocks and A/B test flags.
- Researchers found a leak inside the SSE stream and studied hundreds of SSE dumps collected from the US, Europe and Asia.
- Numbers kept shifting because ChatGPT runs many A/B tests; real values for ANN, BM25, recency and RRF came from a single dump, with broader patterns drawn from hundreds.
- OpenAI patched the leak; the talk covers the roughly two-week window before the fix.
More from Safety
- Podcast: will Texas's new data center rules actually break AI? — thursdai_pod · 2026-11-07
- 'What TikTok sees' screenshot reveals an attractiveness score of 84.7 plus age and mood guesses — cyrus_zei · 2026-11-07
- OpenAI and fmf unveil framework to grade when AI jailbreaks enable serious cyber harm — OpenAI · 2026-11-07
- Beyond prompt injection: a field guide to 20 open-source AI agent security tools — MaryamMiradi · 2026-11-07
- If AI bug-finding scales linearly with compute, defense is in trouble — kuza55 · 2026-11-07
- Nigerian-American scientist Chinasa T. Okolo named to TIME100 AI list — ChinasaTOkolo · 2026-11-07