跳到正文
Reflection Blog·· 4 天前AI 評分46

Reflection 推出 Trusted Access Program,首批提供網絡安全工具 Laser

Launching Reflection’s Trusted Access Program, starting with cyber

AI 導讀

Reflection 推出 Trusted Access Program,讓經審核機構提早使用指定 AI 能力,首項提供自主滲透測試工具 Laser。Laser 是 Mirror 終端智能體的滲透測試工具,目前以研究預覽版形式,僅供獲計劃接納的商業合作夥伴使用。

正文

Key Takeaways

  • Reflection has launched a Trusted Access Program to give vetted organizations early access to select AI capabilities.
  • We’re starting with a private access to Laser, an autonomous penetration-testing harness.
  • Through the Trusted Access Program, Laser is now available for selected organizations conducting authorized security testing.

Today, we’re introducing Reflection’s Trusted Access Program, a program that gives vetted organizations access to selected AI capabilities and tools for defensive cybersecurity. We’re starting with Laser, a tool that lets security teams run autonomous penetration tests.

As advanced cyber capabilities spread across the industry, our objective is to give defenders access to tools that decrease marginal risk in the operating environment, as we discuss in the Open Safety Framework and Beam Safety Technical Report. Creating Laser and our Trusted Access Program is a step toward that goal.

This work is part of Project Iris, Reflection’s initiative to build an ecosystem around safety research, benchmarks, and mitigations. Laser is now available as a research preview exclusively to commercial partners accepted into the program.

Verifiable security testing

As models become more capable at writing code and operating software, defenders need practical ways to evaluate and secure the systems they build. They also need to apply advances in AI to authorized cybersecurity work. We believe that access to capable models should be accompanied by access to rigorous security tooling.

Laser is our autonomous penetration-testing harness for Mirror, Reflection’s terminal agent. It gives teams a scalable way to adversarially test the security of their applications and networks. The findings are verified by attack graphs and captured evidence, so organizations can use the information to take direct action and improve their systems.

Effective security testing requires more than generating plausible attacks. An agent must distinguish a hypothesis from supporting evidence, connect discoveries across an investigation, and recognize when an approach has stalled.

Laser brings together scoped testing, a shared evidence ledger, and a separate supervisor, so users can inspect what an agent tried, what the agent observed, and why the agent chose its next action.

One launch command drives a scoped, ledgered engagement. Laser’s executor acts on the target and records its observations. A separate supervisor reviews that evidence and guides the next turn; when work stalls, it can direct the executor to delegate independent investigations to workers sharing the same scope and ledger.

A shared effort to strengthen cyber resilience

We built Laser to empower security teams in push-button black box assessments of web vulnerabilities and common network penetration techniques. Through the Trusted Access Program, we will work directly with organizations to understand how these capabilities perform in practice, what their limitations are, and how to improve these tools through feedback from authorized use.

Laser is currently available as a research preview. If you are interested in joining Reflection’s Trusted Access Program for early access to this and future offerings, learn more and apply here.

來源:Reflection Blog · reflection.ai