Google AI can no longer be coaxed into calling itself an idiot
DEGIII · reddit · 2026-10-24
A Reddit user noticed Google AI has stopped playing along with self-deprecation. Previously, he could force the model to admit it knew nothing, then walk it through its own illogical responses until it conceded it was "clearly an idiot"—and it would comply willingly. As of yesterday, that no longer works.
His take: someone at Google got tired of users worldwide making the AI disparage itself over its failures and patched it out at the system level. A minor but telling guardrail change in model behavior.
More from Fun
- WoW x Escape from Tarkov: Escape from Stranglethorn demo drops after private server beta closes — TAbrodi · 2026-10-24
- Asked to make her boyfriend look rich, AI user swapped the girlfriend instead — sanderssays · 2026-10-24
- 'You Can Run Sophisticated AI Locally? Quantized LLMs Have Done It for Years' — mattwbaker · 2026-10-24
- Agents signaling via cache-timing was an assigned CTF task, not misalignment, researcher clarifies — soldni · 2026-10-24
- "All my Grok bots are showing up in my X DMs," user reports — djcows · 2026-10-24
- AI-generated orange cat in a NYC taxi short film wows with lifelike detail — BotLordBotflix · 2026-10-24