362 tracked LLM incidents among 1.2B users: a 0.00003% harm rate?
Dan_Jeffries1 · x · 2026-09-20
Dan Jeffries argues AI safety discourse fixates on hypothetical future harms while ignoring real data: only 362 tracked LLM incidents in 2025 against 1.2 billion users — roughly a 0.000030% rate. He suggests LLMs may be among the safest products in history, mocking calls for urgent protective action. A contrarian take likely to spark debate on how AI harms are tracked and counted.
More from AGI Musings
- Jev founder: all AI models are built for human-in-the-loop, not true automation — hardimanjames · 2026-09-20
- cloneofsimo: doubt 'AGI will solve everything' precisely because experts say it — cloneofsimo · 2026-09-20
- Gary Marcus Pushes Back on Antitrust Suit's 'Slowing AI Harms Subscribers' Claim — GaryMarcus · 2026-09-20
- Alignment joke: if Bryan Johnson really wanted to live forever, he'd solve alignment — EthanJPerez · 2026-09-20
- From Greek athletics to AI: what gets farmed next? — JoeJustice · 2026-09-20
- Nature paper: mammalian brain forms from two structures—hive minds may be real — juanbenet · 2026-09-20