Intruder has launched AI-powered pentesting agents designed to mimic human testers, marking what the company describes as the first step toward continuous, automated security assessments that can keep pace with a threat landscape transformed by artificial intelligence.
The London-based exposure management firm, founded in 2015 by former ethical hacker Chris Wallis, announced the release of its initial pentesting agents on April 30, 2026. These agents actively investigate vulnerability scanner findings identified within Intruder’s platform, using the same methods a human pentester or security expert would employ to validate and assess real-world risk.
Industry need: stretched teams and a shrinking exploit window
The announcement comes as cybersecurity professionals confront a rapidly changing environment. According to Intruder’s own research, published in March 2026 as The Security Middle Child Report, 49 percent of security leaders now cite AI and automation as their top investment priority for 2026. The same survey found that 42 percent of mid-market security teams — defined as those at companies with at least $50m in revenue and between 400 and 6,000 employees — describe themselves as stretched, overwhelmed, or consistently behind.
Andy Hornegold, Intruder’s chief security technologist and a former EU Red Team Operations Lead at Mandiant, said the traditional model of annual or quarterly penetration tests is no longer sufficient. “In the age of AI, where attackers can move faster than ever, the volume of vulnerabilities is growing and exploit windows have shrunk from months to days to hours,” he said. “The old playbook that called for a quarterly or annual pentest has long been unfit for purpose. The state of the threat landscape necessitates a new approach, focused on delivering the depth of a manual pentest, on-demand.”
The research briefing notes that AI is empowering attackers to accelerate exploit development, weaponise vulnerabilities faster, and craft more sophisticated attacks. Time-to-exploit has collapsed, making continuous validation increasingly critical for defenders.
How AI pentesting works: from scanner findings to validated risk
Intruder’s AI agents bridge the gap between broad vulnerability scanning and deep manual penetration testing. Vulnerability scanners, the company explains, provide wide, affordable, and frequent coverage but lack the context to determine whether a flagged issue is actually exploitable or sensitive. Traditional pentests offer that depth but are expensive, infrequent, and require scheduling weeks or months in advance.
The new AI agents interact directly with the target system — sending requests, analysing responses, and probing for exposed data — to build a picture of a vulnerability’s real-world impact. They reduce the investigation phase from hours to minutes, allowing security, IT, and development teams to spend less time triaging false positives and more time fixing genuine problems.
Intruder’s longer-term vision is continuous, AI-powered pentesting and red teaming across web applications, external networks, and internal networks. Full-scale, audit-ready web application pentests are expected by the end of Q2 2026 and are intended to provide compliance evidence for frameworks such as SOC 2, ISO 27001, and PCI DSS.
Investigating specific vulnerability types: injection, client-side attacks, and information disclosure
The initial release equips agents to investigate three categories of issues in particular depth. For each, the AI goes beyond what a traditional scanner can achieve by mimicking the contextual reasoning of a human pentester.
Injection issues: The agent validates injection flaws — vulnerabilities that allow an attacker to manipulate an application’s commands, queries, or instructions to gain unauthorised access. It does so by reproducing the scanner’s finding using a range of injection techniques, including error-based, timing-based, and UNION-based approaches. This confirms whether the flaw is genuinely exploitable and what an attacker could achieve.
Client-side attacks: When a scanner flags a client-side vulnerability such as clickjacking — typically because frame-related headers are missing — the AI agent can determine whether the risk is real. Scanners flag clickjacking whenever those headers are absent, but some pages are deliberately designed to be frameable and pose no actual threat. Traditional scanners cannot make that distinction. Intruder’s AI pentesting agent can, by assessing the context and behaviour of the target page.
Information disclosure: Scanners often alert when configuration details or open cloud storage buckets are exposed to unauthorised users, but they cannot assess whether the exposed data is actually sensitive. Intruder’s AI agent confirms the scanner’s finding, reviews what is exposed, and evaluates how an attacker could use it. If it discovers credentials — such as login details or API keys — the agent attempts to verify whether they are still valid, providing a concrete measure of risk rather than a theoretical alert.
Availability and next steps
Issue-level pentesting investigations are now available to free trial users and Intruder customers on the Cloud, Pro, and Enterprise plans, with AI Pentesting credits included. Additional credits can be purchased. The company expects to release new capabilities regularly over the coming quarter, culminating in full-scale web application pentests that can serve as compliance and audit evidence.
Intruder, which now protects over 3,000 companies worldwide, was selected for GCHQ’s Cyber Accelerator and named the fastest-growing cybersecurity company in the UK on Deloitte’s Tech Fast 50 list in 2023. The company has raised approximately $1.48m in total funding, according to PitchBook, with its seed round secured in April 2015.
