Scale AI
- Type
- AI company
- Lieu
- États-Unis — USA
- Dernière vérification
- 2026-09-22
- Prochaine vérification
- 2027-03-21
Pas encore traduit — affiché en anglais.
Comment les joindre
- Vulnerability reports
- Trust Center
Ce canal est conservé uniquement comme référence. Sa vérification est absente, date de plus de 180 jours, ou la période de contact n'est pas ouverte actuellement ; ne vous y fiez pas avant une nouvelle vérification.
- Homepage
- No frontier safety framework (not a frontier model developer); publishes evaluation methodology such as the AI Risk Matrix
Ce qu'il fait
Safety, Evaluations and Alignment Lab (SEAL) produces third-party model evaluations and red-teaming for other developers; internally a trust/security function runs disclosure.
Évaluation franche
Scale's public safety relevance is as an evaluator of others rather than a developer, so external safety reports about Scale itself are rare. Scale had a significant 2025 incident in which public Google Docs exposed confidential client project details and contractor data, surfaced by journalists rather than through disclosure. Since Meta's investment and the departure of leadership to Meta, SEAL's independence as a third-party evaluator has been publicly questioned — which matters if you are considering Scale as a neutral route for a finding about another lab.
Notes
vulnerabilities@scale.com is security-scoped. Scale is not a route for reporting concerns about frontier models generally, and its evaluator independence is contested post-Meta investment. No safe-harbour language found.