AI Verify

Overview

The toolkit focuses its technical testing on fairness, explainability and robustness, with more than eight tests for traditional AI applications. A separate framework assesses responsible implementation against 11 principles, including transparency, safety, security, data governance and accountability. It aligns with principles from the EU, OECD and Singapore, and maps to NIST AI RMF, its Generative AI Profile, the Hiroshima Process Code of Conduct and ISO/IEC 42001. These give teams reference points for governance work, but AI Verify does not establish ethical standards or guarantee a system is safe or unbiased.

AI Verify supports self-assessment as well as third-party testing, and can combine developers’ and compliance teams’ work into a report. The AI Verify Foundation is a not-for-profit subsidiary wholly owned by Singapore’s Infocomm Media Development Authority. The toolkit is released under the permissive Apache 2.0 license.

Key features

Testing and workflow

Tests can be run through the Portal or command line. Users can upload results, complete process checklists and produce reports, bringing technical and organizational assessment into one workflow. Standard templates set out report layouts, tests and checks, while customized reports allow teams to tailor the output. That flexibility is useful when different stakeholders need a shared record, though it does not turn test results into a safety certification.

Data, integrations and extensions

AIVT 2.0 supports tabular and image data and accepts a model file, preprocessing pipeline or model API as input. Plugins from the Foundation or third parties extend the toolkit. Veritas adds fairness and transparency tests for threshold trade-offs as well as explainability plots; its integration is intended to help financial institutions address common safety-baseline and financial-testing requirements from MAS.

Governance and scope

The framework covers traditional and generative AI governance, but the Testing Toolkit is aimed at traditional machine-learning models. Project Moonshot is the separate tool for LLM applications, using benchmark testing and red teaming, with a Web UI, interactive CLI, library APIs and Web APIs for MLOps integration. Teams seeking one toolkit for technical testing of both traditional models and LLM applications should account for that division.

Pricing

AI Verify is free, with a free plan and no paid plan described. That makes it accessible for organizations exploring testing and governance without a stated subscription charge. It is open source under Apache 2.0, so teams can use and extend the toolkit; the value will depend on whether its supported inputs and test scope match their systems.

Platforms

AI Verify is available through web and API interfaces, on Linux, and for self-hosted deployment. On-premises deployment is also supported, giving organizations deployment choices beyond a hosted web workflow. The Foundation says personal data stored or transmitted electronically is protected with appropriate security technologies.

Who it's for

Data-science and compliance users are the clearest audience, but AI owners, developers, researchers, service providers and companies integrating the toolkit can also use it. It is a good fit for teams that want open-source technical testing paired with governance checklists and reports, particularly when self-hosting or API access matters. It is less suited to teams looking for a defined ethics standard, a guarantee of risk-free results or a single traditional-and-generative AI testing toolkit.

Pros and cons

• Pros: More than eight tests address fairness, explainability and robustness, providing technical assessment across several important model qualities.
• Pros: Portal and CLI workflows, checklists and customizable reports can bring developer and compliance work together.
• Pros: Apache 2.0 licensing, plugins and self-hosted options support adaptation and deployment control.
• Cons: The technical testing toolkit targets traditional AI, while LLM assessment is handled separately by Project Moonshot.
• Cons: The framework does not define ethical standards, and the toolkit cannot certify that systems are free of bias or risk.
• Cons: AIVT 2.0 supports tabular and image data, which may not match teams whose data falls outside those types.

Alternatives

Choose AIGovernr if web-based governance is a priority and a paid Professional plan with website scans, findings, gap analysis, action plans, a governance brief and priority support suits the work. Consider Prufer for API or web access and a free trial, while keeping in mind that its free plan has limited features and quotas, retains audit data for seven days and offers no uptime guarantee or SLA credits.

Trusys AI is another freemium option with API, self-hosted and web platforms; its Starter plan includes one application, unlimited functional and security evaluations, 5,000 metrics per month and 30,000 production spans. Verisum offers a free-forever Explorer plan for one assessment and supports API and web access. For API, self-hosted or web LLM security tooling, Openlayer Guardrails has a free Basic plan with one member, five projects, one inference pipeline per project, 20,000 inference logs per month and 20 tests per project.

Grasp, CognitiveView and Enzai are paid alternatives; their listed platforms are web for each. Browse AI Governance Software for more options in the category.

Verdict

AI Verify is a strong choice for teams that want free, open-source testing of traditional AI models alongside governance checks and reporting. Its scope, extensibility and deployment options make it especially relevant to data-science and compliance workflows. Look elsewhere if you need LLM testing within the same toolkit, a prescribed ethical standard or assurance that assessment results mean a system is safe.