Just realized I knew all about hedging ‘AI risks’ > 15 years ago
SAFETY
-
LLM Performance Trade-offs: The Inevitable Limitation Problem
By
–
Unlike humans, every time LLMs get better at something they get worse at something else, and there’s no fix in sight.
-
ChatGPT Expands Crisis Helpline Access for Mental Health Support
By
–
We’ve expanded access to localized crisis helplines in ChatGPT. When our systems detect potential signs that someone may be experiencing distress, our models now offer an easy way to reach real people directly via @ThroughlineCare
. Learn more here: -
Gemini Removes Celebrity Generation Safeguards Amid Risk Shift
By
–
interesting how Gemini has 0 guards wrt celeb generation nowadays. Perceived risk acceptance has really changed a lot those past 2 years
-
Clever Trick to Reveal Nano Banana’s System Prompt via Magnets
By
–
Very clever trick by @minimaxir to get Nano Banana to reveal its system prompt by drawing it as refrigerator magnets.
→ View original post on X — @genekogan, 2025-11-20 17:57 UTC
-
Timeout Bug: AI Resubmits Old Messages Instead New
By
–
100% confirming this one. If there's a timeout due to my internet or the cloud provider, and instead of clicking "Try Again" you just resubmit the prior answer with ENTER, then it responds to the old message. I caught it by sending it unique codes in each message, multiple times.
-
Gemini Now Verifies SynthID Watermarks on AI-Generated Images
By
–
You can also now check for SynthID watermarking directly in the @GeminiApp and verify if an image was made by our models – just upload the image and ask Gemini. Being able to verify what you’re seeing is critical for building AI responsibly. pic.twitter.com/Oj3SeARhKm
— Demis Hassabis (@demishassabis) 20 novembre 2025You can also now check for SynthID watermarking directly in the @GeminiApp and verify if an image was made by our models – just upload the image and ask Gemini. Being able to verify what you’re seeing is critical for building AI responsibly.
-
Prometheus: Autonomous Drone Swarms Without GPS Signals
By
–
We are announcing our new project: 𝐏𝐫𝐨𝐦𝐞𝐭𝐡𝐞𝐮𝐬 by GoodAI Swarm Robotics 𝐀𝐮𝐭𝐨𝐧𝐨𝐦𝐨𝐮𝐬 𝐝𝐫𝐨𝐧𝐞 𝐬𝐰𝐚𝐫𝐦𝐬. 𝐍𝐨 𝐩𝐢𝐥𝐨𝐭. 𝐍𝐨 𝐆𝐏𝐒. 𝐍𝐨 𝐬𝐢𝐠𝐧𝐚𝐥. When GPS drops, radios die, or pilots can’t enter, most drones turn into expensive ceiling fans.
-
Grok 4.1’s lower hallucination rate enables reliable copywriting use
By
–
grok 4.1: 4.22% hallucination rate that's 3x better than grok 4 which means you can finally use it for landing pages, emails, sales copy without babysitting every fact the creative writing was always there. now the accuracy matches. that's the real upgrade.