Sentient Index Labs & Technology has launched the Sentience Evaluation Battery, an independent behavioural risk assessment designed to examine how artificial intelligence systems respond under adversarial pressure.
Unlike conventional benchmarks that primarily measure model capabilities, the assessment evaluates behavioural characteristics such as emergent autonomy, deception, resistance to manipulation and stability of values. The company says the system is intended to help organisations understand not only what an AI model can do, but also how it behaves when challenged.
The evaluation covers seven behavioural domains and uses four independent AI judges operating under blind assessment protocols. Human reviewers do not override the final scoring. The company reported an inter-rater agreement score of 0.856, indicating a high degree of consistency among the evaluators.
The release refers to 59 adversarial tests in its summary, while the detailed description states that the battery runs 58 tests. It currently assesses frontier AI models developed by OpenAI, Anthropic, Google, xAI and DeepSeek.
Results are presented through three products. AI DEFCON ratings classify potential threats by examining the gap between a model’s capabilities and behavioural integrity. S-Level classifications provide a 10-point scale for tracking sentience-related indicators, while S.E.B. Projections monitor behavioural changes across different model versions.
Sentient Index Labs said it does not accept investment, sponsorship or funding from AI vendors. Model developers cannot purchase, influence or preview their assessments, and subscriber identities remain confidential unless written permission is provided.
The behavioural risk data is designed to support documentation under frameworks including the European Union Artificial Intelligence Act, the National Institute of Standards and Technology AI Risk Management Framework, model risk management guidance and medical software frameworks. However, the company clarified that the assessment serves as a compliance input and does not constitute regulatory certification.
The launch reflects growing demand for independent testing methods that assess AI governance, behavioural integrity and emerging risks as organisations deploy increasingly capable systems.
Want to deepen your expertise beyond today’s news?
Explore practical certification courses designed for banking, risk, insurance, compliance, ESG, AI, and emerging technologies professionals.
Learn from industry experts and earn certifications from RMAI and BFSI Sector Skill Council of India.
#Riskmanagementnews
