Israeli AI Defeats 125,000 Hackers in Elite Competitions for $5,000

An artificial intelligence system developed by Israeli startup Tenzai has defeated 125,000 human participants across six elite cybersecurity hacking competitions, outperforming 99 percent of competitors while operating on a budget of just $5,000.

The AI competed in capture-the-flag contests, annual events that challenge cybersecurity experts to execute attacks ranging from password cracking to complex multi-layer exploits designed to breach defenses protecting a digital target. Tenzai's system, built using advanced models from OpenAI and Anthropic, demonstrated what co-founder Pavel Gurvich described as surprising aptitude at "combining exploits for software vulnerabilities."

The performance marks a departure from previous AI applications in cybersecurity, which have largely focused on defensive postures such as threat detection and vulnerability scanning. Tenzai's success in offensive operations—tasks requiring creative problem-solving and adaptive strategy—suggests AI capabilities in this domain have crossed a threshold that was widely considered years away.

"There's still room at the top for humans," Gurvich said, a qualification that acknowledges the AI did not claim first place in the competitions but rather dominated the broad middle tier of professional participants.

The cost efficiency of the operation carries strategic implications. At $5,000 for six competitions, the per-event expense was under $850, a fraction of the salary cost for a single human competitor and orders of magnitude below the budget required to field a competitive team. This economic dynamic could reshape both the offensive security market and the calculus of state and non-state actors evaluating cyber capabilities.

(The competitions in which Tenzai participated were not identified by name in available reports, nor were the specific vulnerabilities exploited disclosed.)

The result arrives as enterprises face mounting pressure to secure systems against increasingly sophisticated attacks. Separate industry analysis indicates that demand for AI-related insurance products among small businesses is expanding as more organizations integrate the technology into operations, creating new attack surfaces. Meanwhile, healthcare systems and payers are grappling with fragmented AI procurement strategies that have produced what one industry voice described as "a mishmash of vendors" and unclear return on investment.

The offensive capability demonstrated by Tenzai also intensifies an existing tension in the AI sector. OpenAI and Anthropic, whose foundation models powered Tenzai's system, have positioned their platforms as secure, enterprise-grade infrastructure. Anthropic's Cowork and OpenAI's Frontier offerings are designed to operate within organizational boundaries, yet the same underlying technology is now demonstrably effective in adversarial contexts. This dual-use reality complicates the governance frameworks that enterprises and regulators are attempting to construct around AI deployment.