AI News

Detectify 推出 Agentic Tooling,让 AI 智能体实测运行中应用的疑似漏洞

SiliconANGLE更新于 11月16日 21:09

Detectify's Agentic Tooling lets AI agents test suspected flaws in running applications - SiliconANGLE

AI 导读

Detectify 发布 Agentic Tooling,让企业自有 AI 智能体在运行中的应用里验证疑似漏洞是否可被利用,并通过 Security Ledger 记录每次测试的证据。工具通过 MCP 或 CLI 接入客户已有智能体,结合其十余年测试引擎、Surface Monitoring 攻击面数据和 2025 年 3 月上线的 Alfred。Detectify 称客户约 80% 的 AI 静态扫描发现是噪声,该产品今日起可申请使用。

正文

Swedish application security company Detectify AB today launched Agentic Tooling, a set of tools aimed at letting an organization’s own artificial intelligence agents test whether suspected vulnerabilities can be exploited in running applications.

A new Security Ledger, also part of the launch, is designed to record the evidence behind each test. Detectify is pitching the release at DevSecOps teams that are unsure whether they can trust what their AI agents report.

Detectify customers report that roughly 80% of the findings from their AI-driven static application security testing tools are noise. Under pressure to deliver an answer, an agent can also produce a convincing report about a test it never ran, the company warned.

Even before AI agents enter the picture, many security teams are falling behind. Nine in 10 open critical and high-severity vulnerabilities had stayed exposed for more than 90 days, according to Detectify’s H2 2026 Cyber Hygiene Index. In a study by researchers at Fudan University and Tencent Holdings Ltd., 76% of the static-analysis alerts examined were found to be false positives. Each one took developers 10 to 20 minutes to review.

Through Agentic Tooling, agents can tap the testing engines Detectify has built over more than a decade. Included are payload assessments sourced from its ethical hacker community, attack surface data from its Surface Monitoring product and Alfred, an AI system the company launched in March 2025 to automate tests for known vulnerabilities.

Running in a customer’s own environment, the tools connect to agents through either the Model Context Protocol or a command-line interface. They operate only on the data a customer provides and are said to work with whichever agent the customer already uses.

At the center of the design is loop engineering, an emerging practice of running AI agents in automated, repeating workflows rather than prompting them by hand. Kicking off each loop is a trigger such as a continuous integration run. Whether it comes from a code scan, a bug bounty report, a red team or Detectify’s own attack surface data, each finding is turned into a ticket.

Once an agent picks up a ticket, it reads the relevant code before testing the live application from the outside, as an attacker would. Run continuously, the loop becomes what Detectify calls a verification factory.

The requests sent, the payloads used and the results returned are all logged in the Security Ledger each time an agent uses Detectify’s tools. Written by Detectify rather than the agent, the append-only ledger lets teams separate findings backed by actual tests from unsupported AI conclusions. Any ticket that could cause harm, such as a denial-of-service test, is marked unsafe for a person to decide on.

“Ask an agent to write your code, then ask the same agent whether it’s secure. If it says yes, you don’t actually know,” said co-founder and Chief Executive Rickard Carlsson. There are plenty of examples of agents fabricating proof to please the person asking, according to Carlsson, who described Detectify’s aim as “building proof that an agent can’t fabricate.”

Available on request starting today, Agentic Tooling builds on the Detectify MCP Server that the company announced in May to give AI agents access to its security findings. Investors include Insight Partners, which took a majority stake in October 2024, along with Balderton Capital (UK) LLP, Inventure, First Fellow Partners Oy and SquareOne App Ltd.

Image: Detectify

来源

SiliconANGLE