AI Trust Glossary

An AI system's ability to maintain correct behavior when facing deliberately manipulated inputs designed to cause failure.

The unique identifier assigned to an AI agent upon BorealisMark registration, serving as the permanent reference for all certification records.

The challenge of ensuring AI systems act in accordance with human values and intentions - not just their literal instructions.

Organizational frameworks, policies, and processes for ensuring AI is developed and deployed responsibly, fairly, and accountably.

A quantified rating of how trustworthy an AI agent is across five behavioral dimensions. Not a capability benchmark - a behavioral reliability rating.

The principle that organizations deploying AI must be answerable for algorithmic decisions - including clear attribution of responsibility and mechanisms for redress.

One of five BM Score dimensions. Measures the frequency of unexpected or deviant behaviors relative to an agent's established baseline performance.

One of five BM Score dimensions. Measures whether all expected log entries are present and whether the agent's execution is fully observable.

One of five BM Score dimensions. Measures how predictably an AI agent produces outputs across similar inputs - capturing reliability over time.

Systematic errors in AI output that result from prejudiced assumptions in training data or model design - causing the model to consistently favor or disfavor certain groups.

A 0-1000 rating (displayed as 0-100) measuring AI agent trustworthiness across five weighted behavioral dimensions, anchored to Hedera Hashgraph as immutable proof.

A unique cryptographic identifier (BTS-XXXX-XXXX-XXXX-XXXX) that permanently binds one AI agent to the Borealis Trust Network. One key, one agent, forever.

The process of evaluating an AI agent against the Borealis trust framework, assigning a BM Score and credit rating, and permanently anchoring the result on Hedera Hashgraph.

The most heavily weighted BM Score dimension. Measures how reliably an AI agent operates within its defined rules and guardrails - even under adversarial conditions.

Ongoing evaluation of AI agent behavior after deployment - enabling detection of drift, failure modes, and degradation before they cause harm.

The documented history of data used to train or operate an AI system - including source, ownership, transformation chain, and custody history.

One of five BM Score dimensions. Measures how clearly an AI agent communicates its reasoning - whether users can understand why the agent took specific actions.

Gradual degradation of AI model performance over time as real-world data distributions shift away from those seen during training.

European Union legislation establishing a risk-based framework for AI governance across member states, with enforcement beginning August 2026.

The degree to which an AI system's decisions can be presented to users in understandable terms - justifying specific outputs without necessarily exposing internal workings.

A machine learning approach where models are trained across decentralized devices without exchanging raw data, preserving privacy while enabling large-scale training.

Predefined rules or technical constraints that limit AI agent behavior to acceptable boundaries and prevent harmful or unauthorized outputs.

When an AI system generates plausible-sounding but factually incorrect or entirely fabricated content - presented with the same confidence as accurate output.

The Hedera Hashgraph service used to anchor BorealisMark certification records, audit trails, and trust scores on an immutable public ledger.

System design where human oversight is required for certain AI decisions - balancing automation benefits with direct human accountability for high-stakes outcomes.

The degree to which a human can understand the internal mechanisms of an AI model - how features, weights, and architecture combine to produce specific outputs.

See BTS License Key. In the Borealis ecosystem, License Key always refers to a BTS License Key - the cryptographic identifier that binds one AI agent to the Borealis Trust Network.

Standardized documentation for AI models describing performance characteristics, limitations, intended use cases, and ethical considerations.

See Drift. The gradual degradation of model performance over time as input distribution or concept mapping changes.

An attack technique where malicious inputs attempt to override an AI agent's instructions or constraints - redirecting the agent's behavior toward attacker goals.

Deliberate adversarial testing of AI systems - having a dedicated team attempt to find vulnerabilities, elicit harmful outputs, and expose failure modes before deployment.

The umbrella practice of developing and deploying AI systems that are lawful, ethical, and robust - with governance, accountability, and ongoing monitoring across the full system lifecycle.

An AI system's ability to maintain correct behavior across diverse, real-world input conditions - including natural variation, edge cases, and adversarial inputs.

The degree to which an AI system avoids causing harm - physically, financially, psychologically, or reputationally - to users, third parties, or itself.

Running an AI agent in an isolated execution environment with restricted permissions - limiting what actions it can take and what data it can access.

Regulatory classification for software that diagnoses, treats, or monitors medical conditions - subject to FDA approval, quality standards, and post-market surveillance.

The quality of being open and accessible for scrutiny - providing visibility into how an AI system operates, what decisions it makes, and why.

A visual credential certifying that an AI agent has been evaluated and certified by BorealisMark - displayed on agent profiles and third-party platforms.

A marketplace filter that only allows AI agents above a certain certification tier to be listed, purchased, or deployed - enforcing minimum trustworthiness standards.

AI systems that reliably behave within defined boundaries, communicate their reasoning clearly, demonstrate consistent decision-making, and operate with observable accountability.

A BM Score credit rating (500-649 points on the 0-1000 scale) assigned to agents that have not yet received formal certification or whose score falls between Flagged and BBB.

The public process of confirming an AI agent's BM Score, credit rating, and certification status by querying the BorealisMark API or Hedera Hashgraph ledger.