IRONSCALES gets Anthropic verification for email defence

IRONSCALES has become the first email security vendor verified under Anthropic's Cyber Verification Program, giving it access to AI functions restricted for most users.

Anthropic's programme is intended for cybersecurity groups that carry out offensive security work, including vulnerability exploitation, penetration testing and red-teaming. IRONSCALES plans to use that access for adversarial simulations that examine a customer's organisation in the way an attacker would.

The move highlights a new point of differentiation in the cybersecurity market as major AI model providers tighten controls on advanced systems. Vendors with verified access may be able to test defences and model attacks in ways unavailable to rivals without similar approval.

Focused on email security, IRONSCALES said the verification will feed directly into its simulation and training systems. These are used to build phishing exercises based on reconnaissance about individual organisations, test detection models against AI-generated attacks and train employees on the fraud tactics most likely to target them.

IRONSCALES says it protects more than 17,000 organisations worldwide. It adds that every simulation, threat detection event and employee response feeds into a broader intelligence network of more than 35,000 security professionals across those customers.

Market divide

The broader significance lies in how frontier AI developers are beginning to influence vendor selection in security software. As access to advanced models becomes subject to formal vetting, chief information security officers may increasingly ask whether suppliers are approved to use those systems for adversarial defence work.

That could create a divide between verified and unverified providers, particularly in email security, where phishing, business email compromise and impersonation attacks are becoming more automated. Companies without this access may still rely on open models or more constrained commercial tools to simulate modern threats.

IRONSCALES argues that attackers often use open-source or uncensored AI tools to create phishing lures, false business pretexts and deepfake impersonation campaigns. By contrast, verified defenders can use more advanced frontier models within a controlled framework designed for legitimate cybersecurity operations.

“We founded IRONSCALES on a simple premise: defenders should always be one step ahead. For years, the security industry accepted that attackers would adopt new technology first and defenders would react. That era is over. Anthropic's verification confirms what we've been building toward since 2014 - a platform where AI doesn't just detect threats after they arrive, it anticipates and preempts them before the first email is ever sent. When a frontier AI lab formally verifies your organization as a trusted defender, that is the market telling you the approach is right,” said Eyal Benishti, founder and chief executive officer of IRONSCALES.

Email focus

The company's business centres on securing Microsoft 365 and Google Workspace environments through API integrations rather than mail-routing changes. Its platform also includes phishing simulation, security awareness training, DMARC management and response automation.

According to IRONSCALES, the verification strengthens the offensive testing side of that offering. Its adversarial simulation engine can now study a customer's organisational structure and likely attack surface more closely, then use those findings to generate tailored phishing and business email compromise scenarios.

This, the company says, moves it beyond generic training templates and allows it to produce exercises based on genuine reconnaissance. The same process can also be used to stress-test detection models against campaigns that more closely resemble real attacks.

For customers, the practical effect may show up in employee training and response speed. IRONSCALES says its automated remediation tools handle more than 99% of threats without human intervention and cut incident response time from 30 minutes to 30 seconds.

It also says that for every 100 mailboxes, its platform catches an average of 67.5 phishing attacks each month that bypass Microsoft 365, Google Workspace and legacy secure email gateways.

Threat shift

Security vendors are under pressure to respond to a rise in AI-assisted fraud, especially attacks that combine email with messaging platforms, voice cloning and video impersonation. The trend has challenged older email filtering products built mainly to detect malicious links and attachments.

Steve Malone outlined how IRONSCALES views that shift.

“The timing is not coincidental,” said Steve Malone, chief strategy officer of IRONSCALES. “Email threats have entered the Phishing 3.0 era: AI-generated attacks, deepfake voice and video impersonation, and multi-channel campaigns that blend email, collaboration tools, and real-time communications. Legacy SEGs were built for an earlier era of bad links and malicious attachments. Behavioural AI vendors are still catching up to social engineering. IRONSCALES has been pioneering the behavioral approach to email protection since 2014. We built the largest crowdsourced threat intelligence network in email security, and launched agentic AI capabilities. CVP verification is the final piece: frontier AI access that ensures the platform stays ahead of threat actors regardless of how fast AI-powered attacks evolve.”

IRONSCALES says its customers include Veeam, Chubb, Loews Hotels and the City of Memphis.