Arnica Sells Proof Of Exploitable Flaws. Its CEO Says Cyber Insurers Should Weigh Theory Too.

Estimated reading time: 4 minutes

A scanner flag marks a flaw an attacker might use. Arnica’s new AI pen tester attacks live applications to show which ones they can. CEO Nir Valtman still wants insurers to weigh both, “at a specific point in time.” He also explains the guardrails on live tests and where Arnica fits in a broader security program.

AI coding tools now change applications daily, Arnica says. Pen testing often still runs by hand, at fixed intervals. Between tests, scanners pile up findings that may or may not matter.

Some policies put a clock on that pile. Chubb’s Neglected Software Exploit Endorsement gives policyholders 45 days to patch a published flaw, CINI reported. After that, more of the loss shifts to the policyholder in stages.

Flags can also shrink under testing. In a report this month, Black Kite flagged one flaw at 2,902 AI suppliers by software version. When it tested, it confirmed 963, CINI reported.

Arnica built its Agentic Pen Testing to address that gap. AI agents attack a company’s live applications the way an attacker would. They try to prove which suspected flaws can be exploited. Proven findings go to developers with proposed fixes. The agents retest once the code changes.

“Penetration testing was built for a world where code shipped in slow iterations,” said Valtman, Arnica’s co-founder and CEO. The Atlanta company began a controlled rollout on Nov. 5.

Proof And Timing In Pen Testing

Valtman sells proof. He wants insurers to weigh it alongside theory. “I think it needs to be a combination of both and at a specific point in time,” he said.

See also  Spektrum Integrates Limit Marketplace Into Cyber Insurance Policy Workflow

Budget is one reason. A pen test is only as good as the time and money behind it, Valtman said. A real risk in the code can go unexplored when either runs short. Arnica pairs its code findings with its pen test agents to aim at the biggest impact.

Time is the other. A third-party package can be secure one day and exploitable once a flaw in it is published, he said. That is why System and Organization Controls 2 (SOC 2) audits observe a company over a period.

What Keeps An AI Pen Test In Bounds

Proof comes from attacking a live application. A test that slips its limits could disrupt production or touch customer data.

Each test runs in a sandbox on Arnica’s side, inside guardrails written into a playbook, Valtman said. Customers can extend the playbooks with instructions of their own.

Arnica uses zero data retention models, he said. Production data does not train them. Teams can also test earlier in development, where there is typically no customer data.

Valtman did not say who carries the loss if a test causes damage anyway.

Pen Test Results On Insurance Applications

Valtman casts Arnica as one layer among many. It “plays a role in the broader cybersecurity program,” he said.

“Most of our midsize and enterprise customers are insured, and we fill that gap,” he added. “There is no other product running side by side to address this requirement by insurers.”

Valtman did not say whether any insurer has accepted Arnica’s results on an application.

How Many Pen Test Findings Stand Up

One number would make the sales case: the share of flagged findings that proved exploitable. Arnica declined to give it. The rate moves with the limits set on each test, Valtman said. A single figure “wouldn’t be fair,” he said.

See also  AI Risk Reshapes Cyber Insurance: Key Takeaways From Lockton Re and Armilla’s “Ready or Not”

Arnica has not said how many organizations are in the controlled rollout.

The closest Valtman came: findings Arnica rates high or critical “tend to have a higher rate of exploitability, as expected.”

FAQ – Agentic Pen Testing

What is agentic pen testing?

AI agents test a live application the way an attacker would. They try to prove which suspected flaws can be exploited.

What did Arnica launch?

Arnica began a controlled rollout of Agentic Pen Testing on Nov. 5. It sends validated findings to developers with proposed fixes and retests after changes.

Why does exploitability matter for patch timing?

Some policies tie cover to patch timing. Chubb’s endorsement gives a 45-day grace period. Arnica’s CEO says insurers should weigh proven and theoretical flaws at a point in time.

How does Arnica protect production data during a test?

Arnica says tests run in a sandbox on its side, with guardrails in the playbook. It says it uses zero data retention models and does not train on production data.

Has any insurer accepted Arnica’s results?

Arnica did not say. Its CEO said most midsize and enterprise customers are insured.

What share of findings are exploitable?

Arnica declined to give a number. It said the rate varies with the limits placed on each test.

Leave a Comment

×