Sovra Recon·Guide · Market

AI Pentest in France: An Overview of Key Players (2026)

By Solan Barlet, co-founder of Sovra Recon
Updated October 29, 2026

In short

In France, AI pentests were offered by about ten providers by the end of 2026: startups specializing in agentic pentests (Escape, Fleuret AI, Sovra Recon), offensive security platforms originating from bug bounty programs (YesWeHack, Yogosha), providers of automated penetration testing or PTaaS (Patrowl, Hydra), and attack validation platforms, both French and foreign (Filigran, Horizon3.ai, Pentera). They differ primarily in the scope of testing, the degree of automation, and the business model (one-time audit, ongoing subscription, or platform). No major audit firm was yet offering AI pentest services at that time.

A market that emerged in 2025–2026

AI pentest has evolved from a prototype to a commercial offering in just a few months. Several indicators illustrate this:

  • Launches: YesWeHack launched its Agentic Pentest offering on June 25, 2026; Pentera announced an AI pentest for web applications in July 2026.
  • Funding: Escape announced an $18 million Series A round in March 2026, and Fleuret AI raised €4 million in October 2026.
  • Ecosystem: According to the Wavestone and Bpifrance French Cybersecurity Startup Radar (June 2026), 70% of French cybersecurity startups incorporate AI into their offerings, compared to 53% a year earlier.
  • Threat: In November 2025, Anthropic described an espionage campaign in which AI carried out 80 to 90 percent of the offensive operations. Attackers are automating; defenders are testing more frequently.

AI pentest providers in France

PlayerOriginCategoryPositioning (according to the provider)
EscapeFranceAgentic pentestOffensive security platform for engineering teams: AI agents that discover, test, and help fix vulnerabilities, with a focus on business logic.
Fleuret AIFrance (Paris)Agentic pentestAutomated AI pentests of web applications and APIs, with replayable evidence; hosting announced in Europe.
Sovra ReconFrance (Marseille)Agentic pentestFully automated audits performed by offensive AI agents on web applications and APIs, either as one-time assessments (RAID) or on an ongoing basis (SENTINELLE); each identified vulnerability is verified.
YesWeHackFrance (Paris)Offensive platform · agentic pentestA platform originating from the bug bounty program, which has offered on-demand AI pentests since June 2026—conducted by AI agents on the attack surface, with optional human triage.
YogoshaFrance (Paris)Bug bounty · PTaaSAn offensive security platform supported by a community of researchers, featuring AI components for triaging and qualifying vulnerabilities.
PatrowlFranceAutomated Penetration Testing · PTaaSAutomated and continuous penetration testing of the attack surface, combining automation with validation of each vulnerability by a penetration tester.
HydraFrance (Gardanne)PTaaSPenetration tests of SaaS websites and APIs: AI agents for recognition and verification, validated by an expert.
Filigran (OpenAEV)France (Paris)Attack ValidationAn open-source exposure management platform; its OpenAEV module simulates attacks, including a mode where an AI agent constructs attack chains.
Horizon3.ai (NodeZero)United StatesAutonomous Penetration TestingAutonomous infrastructure penetration testing (internal, external, cloud), distributed in France by MIEL since 2025.
PenteraUnited States / IsraelAttack ValidationAutomated security validation through controlled real-world attacks, featuring an agentic AI interface and a native AI-powered web AI pentest module.

This list was compiled from the providers’ websites and industry publications, and is limited to offerings with verifiable operations in France. Other international providers offer AI pentests but have no documented presence in France as of this writing.

Four categories of offerings

  • Agentic pentests: AI agents explore and attack an application like a penetration tester, validating vulnerabilities by exploiting them. Suitable for web applications, APIs, and business logic.
  • PTaaS and automated penetration testing: recurring tests of the attack surface, delivered via a platform, often with human validation of each result.
  • Offensive platforms stemming from bug bounty programs: a community of researchers, supplemented by AI agents for triage or on-demand AI pentests.
  • Attack validation (BAS, CTEM): attack scenarios replayed on the infrastructure to verify that defenses hold up; more focused on networks, Active Directory, and the cloud than on application logic.

How to Compare the Services

CriteriaQuestion to ask
ScopeWeb applications and APIs, internal infrastructure, the cloud, mobile: what is actually being tested?
EvidenceIs each vulnerability demonstrated through a reproducible exploit, or is it merely flagged as a potential vulnerability?
RetestDoes each new test cover the entire scope, or only what has changed?
Business LogicDoes the tool test access control between users (IDOR) and business rules, or does it primarily test for known vulnerabilities?
Legal FrameworkWritten engagement letter, signed scope of work, non-destructive testing?
DataWhere are the models and evidence hosted, and how long are they retained?
ModelOne-time audit, ongoing subscription, or self-service platform?
ComplianceIs the report applicable to NIS2, ISO 27001, DORA, or an insurer?

How the Regulations Are Changing Things

The demand for regular testing is driven by several European regulations. DORA has applied to the financial sector since January 17, 2025, and requires certain entities to conduct threat-based penetration tests (TLPT), as governed by Delegated Regulation (EU) 2025/1190. NIS2 requires an assessment of the effectiveness of security measures; in France, its transposition into national law via the Resilience Act was still pending as of late October 2026, following the Senate’s adoption of the bill in March 2025.

Frequently asked questions

Which companies offer AI pentests in France?

By the end of 2026, notable players included Escape, Fleuret AI, Sovra Recon, and Hydra (agentic pentests or PTaaS using AI agents); YesWeHack and Yogosha (offensive platforms with AI components); Patrowl (continuous automated penetration testing); as well as Filigran, Horizon3.ai, and Pentera for attack validation.

Do traditional audit firms offer AI pentests?

As of the date of this overview, no major French audit firm was offering AI pentest services, even though many use automated tools internally.

Agentic pentests or validation platform: which should you choose?

Agentic pentests primarily target web applications, APIs, and business logic; validation platforms (BAS, CTEM) tend to test infrastructure and defenses. The two are complementary.

Is this overview comprehensive?

No: it lists providers whose operations in France can be verified using public sources. The market is evolving rapidly; the page is updated regularly.

Sources

Find out where an attacker would get in.

Our AI agents actually attack your scope, fully automated, and hand you the proof of every vulnerability.