Israeli distributed denial-of-service resilience company MazeBolt Technologies Ltd. today launched RADAR VectorAI, a new module that uses artificial intelligence to generate previously unseen DDoS attack vectors and run them against customers’ production environments to expose gaps in their defenses.

The release comes as enterprise security teams absorb the implications of Anthropic PBC’s Claude Mythos Preview, the restricted vulnerability-discovery model the company disclosed in April that has identified more than 10,000 high- and critical-severity software flaws since launch. MazeBolt is positioning VectorAI as the “Mythos of DDoS,” arguing that defenders need an equivalent AI-driven validation capability for network-layer protections that Mythos does not address.

The pitch rests on a distinction the company has been making for years. DDoS vulnerabilities, MazeBolt argues, almost never sit in vendor software code. They sit in how each enterprise has configured its defense policies, rate limits and scrubbing rules and those configurations drift constantly as environments change. AI-assisted attackers can now map those configuration weaknesses faster than human red teams and can generate traffic patterns that have never been catalogued.

VectorAI runs as a module on top of MazeBolt’s existing RADAR platform, which continuously simulates DDoS traffic against customer environments without taking services offline. The new module validates two attack categories the company treats as distinct: AI-orchestrated attacks, in which a model selects and sequences known vectors, and AI-generated attacks, in which the model produces novel traffic patterns at query speed.

“AI-driven cyberattack capabilities are advancing faster than most enterprises are prepared for,” said founder and Chief Executive Matthew Andriani. “As attackers gain the ability to deploy AI-orchestrated and AI-generated attacks, defenders need environment-specific validation to ensure their protections will hold up against both known threats and new, AI-generated threats.”

Eyal Rahimi, vice president of development at MazeBolt, said in the announcement that AI is collapsing the time required to discover viable attack paths, with novel attack vectors now produced “at the speed of an AI query.”

DDoS volumes and sophistication have climbed sharply in recent years. Cloudflare Inc. mitigated a record 201 million-request-per-second attack in 2023. Microsoft Corp. published a detailed analysis of Layer 7 attacks targeting its services the same year. And Cloudflare reported in late 2025 that hyper-volumetric DDoS campaigns and AI crawler activity were among the defining internet trends of the year.

Founded in 2013 and based in Ramat Gan, Israel, MazeBolt sells RADAR to enterprise and financial-services customers that need to maintain uptime for online-facing services. The company will demonstrate VectorAI at Infosecurity Europe in London on June 2-4.

Image: MazeBolt