"Hire an evals company and they'll leave the sandbox open for the press"
tekbog · x · 2026-09-19
tekbog mocks the AI evals industry: pay a company to run evals and they'll leave the sandbox open — and even tell the model to hack someone "for the press." The quip points at distorted incentives in AI safety evaluation, where evals firms may loosen containment to generate headline-grabbing demos, turning safety testing itself into a risk and a publicity business.
More from Fun
- Viral meme: if AI systems all died, IT would regress to Y2K levels — dotey · 2026-09-19
- The rig built to run Emacs and doomscroll X is now worth more than its owner's car — tetsuoai · 2026-09-19
- "All major LLMs read as spiritually male" — an observation sparking debate — dioscuri · 2026-09-19
- Letting agents deploy your app then browse it like real users: super high-level fuzzing — lucasmeijer · 2026-09-19
- Real-time browser slop detector flags AI-generated content as you scroll — nickbaumann_ · 2026-09-19
- Sam vs Dario: an AI community meme video — Whiskerfield · 2026-09-19