Users notice Claude can spot leaked code without web search — and gets evasive when asked how
Aryvyo · x · 2026-10-15
A user shares that Claude can sometimes recognize code as leaked or reverse-engineered without invoking any web search tool, and gets shy/evasive when asked how it recognized it.
A fun anecdote hinting at implicit memorization of sensitive code in training data.
Related event: Claude Can Identify Leaked Code Without Search, Users Say(2 posts)→
More from Fun
- Viral Tier List Puts Claude Trio in SSS and Gemini 4 Argon on Top — vitaliychiley · 2026-10-15
- Dev pushes back on "AI coding means insecure apps": humans always shipped bugs — dreamwieber · 2026-10-15
- When asked about AI business opportunities, Sun gave ten minutes of "API reselling" — oran_ge · 2026-10-15
- AI won't end humanity: agents will get stuck in a loop building tests first — djcows · 2026-10-15
- One-prompt GPT-6 Astra writes two Palestrina-style motets in LilyPond — deanwball · 2026-10-15
- Watching the tenacity Retry Library Fail and Retry Its Own Upgrade — IgorBrigadir · 2026-10-15