Vibeleaderboard
← All Intel
Intel / article

Launching Trusted Access Program

Source
reflection.ai
Author
Reflection AI editorial sitemap
Date

Context

Reflection, the company behind the model Beam, has started a Trusted Access Program that gives vetted organizations early access to selected AI tools for defensive cybersecurity. The first tool is Laser, which runs autonomous penetration tests: authorized, simulated attacks on an organization's own applications and networks to find weaknesses before real attackers do. Laser is a research preview, available only to commercial partners Reflection accepts into the program for authorized security testing. Reflection's stated reason for the gate is that advanced cyber capabilities are spreading across the industry, and it wants defenders to have tools that lower the overall risk in that environment.

Laser is a , the software that directs an , built for Mirror, Reflection's terminal agent. One command starts an engagement limited to a defined scope. An executor agent acts on the target and writes what it observes into a shared evidence ledger, a running record of the investigation. A separate supervisor reads that record and decides the next step, and when progress stalls it can have the executor hand independent lines of investigation to further workers that share the same scope and ledger. Reflection says findings are backed by attack graphs, which map how individual steps chain into a working path into a system, and by captured evidence. Reflection says the design lets a security team inspect what the agent tried, what it saw and why it chose its next move, because useful testing has to separate a hypothesis from the evidence that supports it rather than just generate plausible attacks.

Reflection describes Laser as built for push-button black-box assessments, meaning tests run from the outside without access to source code, covering web vulnerabilities and common network penetration techniques. The announcement publishes no results on how well Laser finds real vulnerabilities. Reflection says it will work directly with participating organizations to learn how the tool performs in practice and what its limitations are, and it gives no eligibility criteria beyond vetting and authorized use.

Key takeaways · AI-distilled
  • Reflection's Laser is an autonomous penetration-testing harness for its Mirror terminal agent, offered only as a research preview to commercial partners accepted into the new Trusted Access Program for authorized security testing.
  • Laser splits the work: an executor acts on the target and records observations in a shared evidence ledger, while a separate supervisor reviews that evidence, guides the next turn and, when work stalls, delegates investigations to workers sharing the same scope and ledger.
  • Reflection says Laser's findings are verified by attack graphs and captured evidence, and that users can inspect what the agent tried, what it observed and why it chose its next action.
  • Laser targets push-button black-box assessments of web vulnerabilities and common network penetration techniques. Reflection says it will work with program members to learn how the tool performs in practice and where its limits are.
Terms in this piece · Glossary
  • open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
  • agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
  • AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Read the source reflection.ai
More from Reflection AI editorial sitemap
Recommended reads
Comments

Checking sign-in…

Loading comments…