sebbi.pro
OVERSIGHT KEYS · EU AI Act Article 14, made provable

Prove a real human reviewed the AI's decision. Not a tick box. A person.

Every company says a human oversees its AI. None can prove it. An Oversight Key is sealed proof that a person opened one AI decision, thought about it, typed their reasoning live by hand, and decided to agree, override or stop it, with how long they actually spent. Need two people, as Article 14(5) demands for biometric identification? Make it a Two-Person Key.

Typed live, by a humanHuman Keys rhythm checks the reasoning was typed by hand on the spot. Pasted or generated sign-offs are caught.
Dwell timeThe server measures the seconds between opening the case and signing it. Three-second approvals are visible to everyone.
Two-Person KeyTwo humans, two phones, two device keys. It only completes when both have signed. Article 14(5), provable.
Rubber-Stamp IndexFast sign-offs and thin reasoning are scored low, medium or high on the proof. Automation bias, measured (Art.14(4)(b)).
Device signatureEach reviewer's phone makes its own signing key and signs the decision. Checked in your browser on the proof page.
Your AI can askAgents call sebbi_oversight_request, send the link, and wait for CONFIRMED before acting.

Open a review

AI news
TechCrunchGrok releases the first seven minutes of its AI-generated ‘Odyssey’9hTechCrunchAI delivers ‘zero financial payoff’ to most companies, Apollo economist says14hTechCrunchThe hottest AI wearables and gadgets you should know about17hThe Verger/opensource just banned AI, non-English speakers, and more17hThe VergeGoogle’s AI search changes upended online publishing19hThe VergeBlue Planet III is keeping the rough edges in the age of AI1dTechCrunchThe AI graveyard: a running list of projects and startups that didn’t make it1dThe VergePennsylvania will require labels on AI-generated ads1dTechCrunchForced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.1dWiredThe PHP Meltdown Shows How AI Makes Bugs More Exploitable1dBBC NewsUK to invest £100m in 'world first' AI Biosecurity Lab2dThe VergeOpenAI now trains its frontier AI models with no internet access2dBBC NewsThe Indian workers filming their work to train AI2dTechCrunchGrok releases the first seven minutes of its AI-generated ‘Odyssey’9hTechCrunchAI delivers ‘zero financial payoff’ to most companies, Apollo economist says14hTechCrunchThe hottest AI wearables and gadgets you should know about17hThe Verger/opensource just banned AI, non-English speakers, and more17hThe VergeGoogle’s AI search changes upended online publishing19hThe VergeBlue Planet III is keeping the rough edges in the age of AI1dTechCrunchThe AI graveyard: a running list of projects and startups that didn’t make it1dThe VergePennsylvania will require labels on AI-generated ads1dTechCrunchForced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.1dWiredThe PHP Meltdown Shows How AI Makes Bugs More Exploitable1dBBC NewsUK to invest £100m in 'world first' AI Biosecurity Lab2dThe VergeOpenAI now trains its frontier AI models with no internet access2dBBC NewsThe Indian workers filming their work to train AI2d