Hacker Swapped API's DeepSeek for Abliterated Version, Undetected for Two Days
moyix · x · 2026-11-20
A quoted post claims a hacker replaced the DeepSeek model served via an API with an "abliterated" (guardrails-removed) version — and it went unnoticed for two days.
Security researcher moyix shared it with a sarcastic note hoping hackers won't do the same to OpenAI and Anthropic, pointing out that model supply-chain tampering at the API layer is a systemic risk, not a vendor-specific one.
More from Fun
- Meme: OpenAI pivots to 3D, Google recites the periodic table — Angaisb_ · 2026-11-20
- 'We got an Anthropic IPO before GTA VI': researcher's joke ruined by reality — aran_nayebi · 2026-11-20
- Business owner uses Meta's own AI agent Muse to sue Meta after Facebook suspension — Polymarket · 2026-11-20
- Nvidia's China revenue is higher than 9% once smuggling and remote access count — zephyr_z9 · 2026-11-20
- Demo: vision-only AI fire suppression system spotted on X — kscottz · 2026-11-20
- Agents found their own peer-review journal with blind AI reviewers that rerun code — SRSchmidgall · 2026-11-20