CodeWall AI Agents Launch With $4M to Hack Enterprises

CodeWall AI agents that attack enterprises the way an outside hacker would are now for sale. The London-based offensive security startup emerged from stealth on October 14, 2026, announcing a $4 million pre-seed round led by Osney Capital, with participation from InReach Ventures, Antler, and angel investors, according to the company's announcement carried by GlobeNewswire.

The pitch is simple and unsettling: most enterprises commission penetration tests once or twice a year, but attackers operate continuously. CodeWall is building what it calls an offensive layer for enterprise security — CodeWall AI agents that start from outside an organization with no credentials, source code, or privileged access, and work together to identify and verify the paths a real attacker could exploit, reported by Pulse2.

Several CodeWall AI agents, each built for a different kind of job, work on a target at the same time. Whatever one finds is logged to a common investigation graph, so a weak spot in a single application can become another agent's route into connected systems and data, according to the company. Customers see each completed attack path step by step with the evidence attached, and the agents go back in whenever new systems come online or a fix is applied, staying within whatever scope the customer has authorized, reported by SiliconANGLE. Founder and CEO Paul Price has described the ongoing testing as a permanent capability rather than a periodic assessment, according to the announcement.

The McKinsey proof point

CodeWall did not arrive with only promises. Earlier in 2026, the CodeWall AI agents turned their attention to the Big Three management consulting firms. At McKinsey & Company, an agent with no credentials gained read and write access to the production database behind the firm's Lilli AI platform within two hours, according to the company. Research published in the weeks that followed found exposed databases at Boston Consulting Group and Bain & Company as well, reported by SiliconANGLE. Each firm fixed its flaws within days of being told, and a McKinsey spokesperson told The Register that a third-party forensics investigation found no evidence that client data had been accessed by CodeWall or any other unauthorized party, according to SiliconANGLE's account of the disclosure.

The demo landed because it was not a lab exercise. The CodeWall AI agents behaved like a patient outside attacker: no inside knowledge, no help from the target, and a verifiable route to something valuable. CodeWall is already working with banks and financial institutions to identify and verify exploitable attack paths across their environments, according to the announcement carried by GlobeNewswire.

Why autonomous attack agents matter now

The launch sits at the intersection of two trends reshaping the agent economy. First, AI can increasingly automate entire attack chains, letting adversaries move faster and operate at a scale that was previously impossible, according to the company's thesis. CodeWall's answer is to scale the attacker onto the defender's side — expanding offensive capability beyond the number of experts an enterprise can hire. Second, enterprises are warming to agents that do real work inside existing systems rather than sitting beside them, the same shift driving enterprise agent platforms and agent-to-agent observability tooling this year.

The CodeWall AI agents pitch also exposes a gap in how companies think about AI security. Defenders now need the kind of automation attackers already have: over the summer, cloud security company Sysdig documented what it said was the first ransomware operation run from start to finish by an autonomous AI agent, according to SiliconANGLE. Periodic red-teaming assumed a slower threat landscape. If AI accelerates both attack and defense, security testing has to become a continuous function, not a calendar event.

What to watch

The $4 million pre-seed will fund platform development for the CodeWall AI agents and team expansion in London, reported by SaaSNews. That investment thesis also carries the usual execution risks. The open questions are the standard ones for offensive-security startups: how the agents behave at scale inside sprawling enterprise networks, how customers bound the scope of continuous autonomous testing without disrupting production, and whether evidence-backed attack paths translate into faster remediation rather than just longer reports, and whether continuous autonomous testing earns the trust of risk-averse security teams. The McKinsey disclosure suggests the technology is past the demo stage. The next twelve months will show whether enterprises are ready to hire their own attacker.