Giskard
Giskard AI SAS
European platform for red teaming, evaluating and guarding AI agents, with an Apache 2.0 open-source library and on-premise deployment
Last checked October 15, 2026Quick facts: Giskard is an AI compliance vendor founded in 2021 and headquartered in Paris, France. The vendor publicly documents coverage for EU AI Act and SOC 2. Pricing is available on request. Profile last checked October 15, 2026; sources are cited at the bottom of this page.
About Giskard
Giskard is a Paris-based company founded in 2021 by Alex Combessie, Jean-Marie John-Mathews and Andrei Avtomonov. Combessie and John-Mathews are now co-CEOs, and Giskard says its co-founders previously built AI systems at Dataiku and Thales. It sells software and services for testing and securing LLM agents. Its first funding round, €1.5 million led by Elaia with Bessemer Venture Partners, was announced in December 2022. Giskard says it also received a €3 million strategic investment through the European Commission's EIC Accelerator in 2023 and a €3 million Bpifrance grant in 2024 for LLM evaluation research with Mistral AI, Artefact, INA and BnF. CORDIS lists a €2.5 million EU contribution to its EIC project, which ran from September 2023 to August 2025. Giskard Hub is the paid platform. Continuous red teaming runs multi-turn attacks that adapt to the agent's answers, generated from the company's own documents and knowledge bases. It tests for hallucinations, stereotypes and discrimination, harmful content, disclosure of personal information and prompt injection. Testing is black-box: the agent only needs an API endpoint, and coverage is limited to text-to-text conversational agents. The Hub also schedules agent evaluations and manages test datasets, custom checks and human review, with alerts, SSO, role-based access and audit logs. It runs as SaaS, in a private cloud or on premises, with EU or US data processing. Giskard Guards, launched in April 2026, is a runtime guardrail layer that runs in the customer's infrastructure. It inspects prompts, responses, tool calls and multi-step reasoning, and its policies are written as code in OPA/Rego. Giskard also sells AI security audits that end in a go/no-go report, and it is offering a free trial of a new Agentic Red Teamer to a limited number of users. The open-source library is Apache 2.0 licensed. Version 3, a rewrite for multi-turn agent testing, became stable in August 2026. Version 2 is no longer actively maintained, and tabular models have no v3 equivalent. Giskard also publishes Phare, an LLM safety benchmark launched in February 2025 with Google DeepMind as research partner. Logos on its site include BNP Paribas, AXA, BPCE, Société Générale, Crédit Agricole, Michelin, Decathlon, L'Oréal and Doctolib. **AI compliance coverage** Guards includes an EU AI Act policy pack for prohibited practices such as social scoring and real-time biometric identification, and Giskard says Guards covers OWASP LLM risks. It doesn't map its products to the NIST AI RMF or ISO 42001. Its trust center, run on Bastion, lists GDPR and a SOC 2 Type 2 report and shows HIPAA as in progress. **Who it's for** Enterprises, especially in European banking and insurance, that run customer-facing LLM agents and want testing and runtime controls they can deploy on premises.
Featured in
Giskard is ranked in the following independent collection.
Frameworks supported
Regulations, voluntary standards and compliance programs Giskard documents support for on their own materials. Chip shading reflects the strength of the claim, not an independent audit.
Regulation · EU · Phased — in force; high-risk rules deferred
Attestations held
Third-party security attestations and certifications Giskard documents on their own materials. These are point-in-time auditor opinions, not regulatory compliance. Always request the current report or certificate directly from the vendor before relying on it.
Attestation · United States (AICPA)
Giskard features
Capabilities Giskard markets publicly. Inclusion means the feature is documented on the vendor's site — not that it's best-in-class. Last checked October 15, 2026.
LLM Red Teaming
Automated adversarial testing of LLMs for jailbreaks, prompt injection, and unsafe outputs.
Agentic Red Teaming
Automated adversarial testing using AI agents to probe model and application weaknesses.
LLM Evaluation
Systematic testing of LLM outputs for correctness, relevance, safety, and consistency using automated scorers, rubrics, or human review.
Hallucination Detection
Automated scoring of LLM outputs for factual accuracy and groundedness, typically via retrieval-backed verification or reference comparison.
Bias & Fairness Testing
Automated statistical testing for disparate impact across protected attributes, with audit-ready reports.
AI Guardrails
Runtime controls that constrain LLM input and output to enforce safety, compliance, and brand policies.
Prompt Injection Defense
Detection and mitigation of malicious prompts intended to bypass system instructions or exfiltrate data.
Data Leakage Prevention
Inspection of LLM outputs to block PII, secrets, and regulated data from leaving the boundary.
Policy Enforcement
Mechanism for applying organizational and regulatory policies to AI usage at runtime.
Human-in-the-Loop
Workflows that require human review or approval for AI decisions.
Audit Logging
Tamper-evident logging of governance events (approvals, model changes, policy decisions) required by EU AI Act Article 12 and similar regulations.
Industries served
Integrations
Documented by Giskard in public product materials.
- LiteLLM
- OpenAI API
- Anthropic API
Giskard pricing
No prices are published. Two plans: Open-source, a free Python library for local agent experiments with community support, and Professional, quoted on request ('Get your AI audited'). Professional adds automated audits with 50+ adversarial probes, evaluation dataset generation, SSO and role-based access, CI/CD integration through the SDK, on-premise, private cloud or SaaS deployment, and support with SLAs. Advisory on fixes and guardrails is an extra service. Giskard advertises a free Agentic Red Teamer trial with limited spots and an offer for startups and small teams, without stating terms for either.
Vendor changelog
Durable events we've independently sourced — acquisitions, certification milestones, product launches. Each entry cites a primary or trusted secondary source.
- product launch
Launched Giskard Guards
On-premise runtime guardrails for AI agents, with policies written in OPA/Rego and an EU AI Act policy pack.
Giskard - product launch
Released open-source v3 in beta
A rewrite of the library for multi-turn agent testing; v3.0.0 was released as stable on August 26, 2026.
Giskard - product launch
Launched the Phare LLM benchmark
An independent benchmark for hallucination, bias, abuse and harmful content, with Google DeepMind as research partner.
Giskard - funding
Started an EIC Accelerator project
Two-year EU-funded project on AI quality assurance; CORDIS lists a €2.5M EU contribution.
CORDIS - funding
Raised a €1.5M first round led by Elaia
With Bessemer Venture Partners and angel investors.
Tech.eu
Pros and cons of Giskard
Pros
- One vendor for the agent lifecycle: continuous multi-turn red teaming, scheduled evaluations and runtime guardrails (Guards).
- Deployment options for regulated buyers: SaaS, private cloud or on premises, with EU or US data processing; Guards runs in your own infrastructure.
- Apache 2.0 open-source library (v3, stable since August 2026) for testing locally before buying the Hub.
- Large European banks and insurers among the logos on its site, and Guards ships an EU AI Act policy pack for prohibited practices.
Cons
- No published prices: Professional is quoted, and the free Agentic Red Teamer trial and startup offer have no stated terms.
- Its homepage and pricing page claim HIPAA compliance, but its trust center lists HIPAA as in progress. It doesn't publish its SOC 2 auditor or report date.
- The open-source v3 is a rewrite: v2 is no longer actively maintained and tabular models have no v3 equivalent, so v2 users face a migration.
- EU AI Act support covers only prohibited practices, through the Guards policy pack. There is no mapping for high-risk obligations, the NIST AI RMF or ISO 42001.
- Continuous red teaming covers conversational agents in text-to-text mode only.
Alternatives to Giskard
Frequently asked
What is Giskard?+
Giskard is a Paris-based company founded in 2021 that tests and secures LLM agents. Its Giskard Hub runs continuous red teaming and evaluations, Giskard Guards adds runtime guardrails, and it maintains an Apache 2.0 open-source testing library.
Is Giskard open source?+
Its testing library is open source under Apache 2.0. Version 3 is a rewrite for multi-turn agent testing and became stable in August 2026; version 2 is no longer actively maintained. Giskard Hub and Giskard Guards are commercial products.
How much does Giskard cost?+
Giskard doesn't publish prices. The open-source library is free; the Professional plan, which includes the Hub, audits, enterprise security and support, is quoted on request. It advertises a limited free trial of its Agentic Red Teamer and an offer for startups and small teams.
Is Giskard SOC 2 and HIPAA compliant?+
Giskard's trust center lists a SOC 2 Type 2 report, available on request, and GDPR. It lists HIPAA as in progress, although its homepage and pricing page list HIPAA compliance. Ask for the SOC 2 report and the HIPAA status before you buy.
Does Giskard support the EU AI Act?+
In part. Giskard Guards includes an EU AI Act policy pack that blocks prohibited practices such as social scoring and real-time biometric identification. Giskard doesn't document support for the Act's high-risk obligations.
Sources
- Giskard: Homepage (Hub, red teaming, evaluation, guardrails, customers, security claims)
- Giskard: Pricing (Open-source and Professional plans)
- Giskard: About (founders, investors, funding history)
- Giskard: Continuous red teaming
- Giskard: Guards (runtime guardrails, EU AI Act policy pack)
- Giskard: Giskard Guards launch (April 30, 2026)
- Giskard Trust Center (GDPR, SOC 2 Type 2 report, HIPAA in progress)
- Giskard: Trust Center, compliance and SLA (July 2026)
- Giskard docs: Giskard Hub
- Giskard docs: Open-source library (v3; v2 migration)
- Giskard: Announcing Giskard open source v3 (April 2, 2026)
- Giskard open-source README (Apache 2.0; v2 no longer actively maintained)
- Release Alert: Giskard-AI/giskard-oss releases (v3.0.0 stable, August 26, 2026; v3.0.1, October 2, 2026)
- Giskard: Phare LLM benchmark launch (February 19, 2025)
- Tech.eu: Giskard raises €1.5M led by Elaia (December 2022)
- Giskard: Giskard closes its first financing round (December 15, 2022; founders)
- Giskard: Giskard integrates with LiteLLM (OpenAI, Anthropic, Mistral and other providers; December 5, 2024)
- CORDIS: Giskard EIC Accelerator project (Paris; €2.5M EU contribution, 2023-2025)
- Journal du Geek: Mistral, Giskard and Artefact consortium selected by Bpifrance (May 2024)
- Tech.eu: Giskard launches Guards (May 2026)
Keep reading
See an error or outdated detail?
Profiles carry a last-checked date. If something is out of date or wrong, send a correction and we will review it.
Work at Giskard?
Claim this listing to propose edits to the tagline, description, pricing notes, and headquarters details. Every change is still reviewed by our editorial team.