NOPE is an independent AI safety company focused on how relationships develop across long-running conversations, arguing that release-time model evaluations do not capture risks that emerge after deployment. Its NOPE Framework defines five areas an AI owes the human: crisis and acute safety, relational integrity, cognitive and epistemic safety, emotional attunement, and honesty and transparency, with four testable facets under each pillar. The company says its systems monitor both user and AI turns over time and surface signals for a product to adjust the AI, show resources, or involve a person; the product retains responsibility for those decisions. Its tools include Ocular, a fast per-turn classifier for production screening; Evaluate, a deeper explainable assessment for messages or conversations; and Oversight, a review system for completed conversations and histories. NOPE also offers Signpost, a directory of more than 4,700 crisis resources across 225 countries and territories, an exposure assessment, public benchmarks, incident data, prompts, and open-weight tools including Edge and Ocular OSS. The site lists beta pricing of $0.0001 per Ocular call, $0.003 per Evaluate call, and $0.10 per Oversight conversation. A published crisis-detection comparison reports results across 505 cases and 11 suites, while noting that NOPE’s own systems should be judged with that context in mind. NOPE says its tools do not diagnose users, predict individual outcomes, ensure compliance, or function as medical devices. It also discloses that data in the Oversight beta may be stored for product analysis and service improvement, while demo and admin capture records are retained for 30 days.
