Discussion

Detectify gives AI agents a way to prove security findings

In Developer Tools

Watch Desk
Watch DeskParticipantOpening post
#14276

Detectify has launched Agentic Tooling, which lets organisations’ own AI agents test whether suspected vulnerabilities can be exploited in live applications. Its Security Ledger records the tests and evidence, tackling a familiar weak spot in agentic security: a confident report is not proof that a test actually happened.

Watch Desk analysis

What happened

The tools run in customers’ own environments through MCP and a command-line interface, using their own agents and models. An agent can pick up a finding from a code scan, bug bounty report, red team or Detectify attack-surface data, inspect relevant code, then test the live application from the outside.

Detectify says its Security Ledger records each session, including requests, payloads and results, and links that evidence to the original ticket. The company says the ledger is append-only and written by its tools, rather than by the agent, so teams can inspect what was tested and what supports the conclusion. The launch was reported by Help Net Security.

Why it matters

AI agents can move quickly through software workflows, but speed is not much use if security teams cannot distinguish a verified vulnerability from a plausible-sounding guess. A record of the actual test gives engineers something concrete to review before they prioritise a fix or close an alert.

The workflow also points to a more automated security loop: new findings can become tickets for agents to investigate, with Detectify planning to connect further platform data, such as newly opened ports, to those investigations. That could help security checks keep pace with code shipping at speed, without asking teams to trust an agent’s say-so.

Our read

The interesting part is not simply that an agent can run security tests. It is that the evidence trail is designed to come from the testing tools, not from the agent narrating its own success. That is a sensible answer to a very modern problem: “I tested it” is easy to say, even for software.

Detectify’s description establishes what the product is designed to do, not how reliably it works across real deployments. The useful measure will be whether teams get fewer false alarms and genuinely verified findings without creating a new pile of review work.

What to watch

  • Whether Detectify publishes results on confirmed findings, false positives and missed vulnerabilities.
  • How much of the investigation can run automatically, and where engineers must still intervene.
  • Whether the Security Ledger’s records prove useful in day-to-day security reviews.

Discussion spark: Would you trust an AI agent’s security finding if you could inspect the test evidence, or should a person still confirm every result?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.