LNX approved for Anthropic's Cyber Verification Program

Anthropic has approved LNX Inc. for Defense Access in its Cyber Verification Program. The approval lets our security team use Claude, including Anthropic's most capable models, for work the public versions refuse by default: investigating incidents, taking apart malware, and confirming whether a reported vulnerability is real on the systems we maintain for clients. It does not change who makes decisions. People still read the alerts and people still call you.

What the program is

Anthropic ships its generally available models with conservative cyber safeguards. Its own description is that they "block most cyber work", because the same capability that helps a defender find and fix a weakness can help an attacker use it. The Cyber Verification Program is how verified security teams get past those blocks for legitimate work.

In October 2026 Anthropic reorganized the program into three tiers and folded its earlier Project Glasswing into it:

• Defense Access is for defensive work: "security operations center and incident response tasks, reverse-engineering malware, and analyzing and validating vulnerabilities." This is the tier LNX holds.
• Red Team Access adds authorized penetration testing and red-team operations against systems the applicant is permitted to test. It is open to organizations only and takes a few weeks to review.
• Specialized Access has the fewest blocks and is reserved for a small set of organizations that test safety-critical systems. Anthropic reviews those applicants with the US government.

Every tier requires verification and proof of the security controls Anthropic specifies for it. Data retention is also required, so Anthropic can monitor for misuse. Those are reasonable terms. They are the same kind of terms we ask of our own clients before we switch on a powerful tool.

What it means for clients

Most of the security work on our cybersecurity page is routine: alerts reviewed around the clock, patches applied, backups tested, evidence collected for the insurer. The hard minutes are the ones after something gets through. An unfamiliar executable on a workstation. A sign-in from a country nobody in the office has visited. A vendor's advisory that may or may not apply to the firmware on your firewall.

Defense Access helps with exactly those minutes. Concretely, we can now use Claude to:

• Analyze suspicious files and scripts found on a client device, to learn what they were built to do before deciding how far to isolate.
• Read attacker tooling recovered during an incident, so containment covers what the attacker actually planted rather than what we guessed.
• Validate vulnerabilities against the systems we manage: whether a published flaw applies to your version, whether the mitigation in place actually closes it, and what the fix order should be.
• Work through investigation evidence faster: sign-in logs, endpoint telemetry, and email headers, during the first hour when speed matters most.

The models are an analyst's tool, not an analyst. Every finding is checked by a person on our team before it changes anything in your environment, and isolation and recovery steps still follow the plan agreed in your service agreement.

What it does not mean

It is worth being precise, because AI claims in security marketing are usually vague.

• It is not a certification. Anthropic verified that LNX is a legitimate security provider doing defensive work. It did not audit our controls on your behalf, and we still do not sign compliance attestations.
• It is not penetration testing. That is the Red Team tier, with its own review. If you want offensive testing of your own systems, we will say so plainly and scope it separately.
• It does not loosen our rules on your data. Our AI security and governance page describes the controls we ask clients to put around AI tools: a business tenant, single sign-on, data rules, and logs. We hold ourselves to the same ones.

Why we applied

We have spent most of 2026 helping Metro Detroit businesses decide how AI tools should be allowed into their companies, and writing a daily threat briefing for managed service providers. Both taught us the same thing: attackers already use these models. Defenders who cannot are working with one hand behind their back. The verification program is Anthropic's answer to that imbalance, and we wanted to be on the right side of it early.

If your business is renewing cyber insurance, has had a close call, or simply wants to know whether its security provider can keep up, book a security review. The review ends in a ranked list with costs, and nothing in it obliges you to buy the rest.