Sweet Security Unveils AI-Powered Sweet Attack to Benchmark and Preempt Advanced Threats

New updates have been reported about Sweet Security.

Sweet Security has launched Sweet Attack, an AI-driven red-team agent designed to help enterprises counter increasingly sophisticated, AI-enabled attackers and to quantify their exposure against benchmarks such as Anthropic’s Mythos. Built on Sweet Security’s existing runtime telemetry, the tool maps real attack paths in live production environments by correlating runtime topology, Layer 7 exposure, deployed code, identity relationships, and application behavior to surface only exploitable chains, not theoretical vulnerabilities.

For CISOs and boards now seeking numeric proof of resilience, Sweet Attack executes simulated attacks at AI speed, logs only paths that can actually be exercised, and produces evidence-backed, prioritized remediation plans that shift security teams from broad patching to targeted breach prevention. Early adopters report that Sweet Attack identified exploitable paths in days that top-tier offensive security firms had missed and provided enough operational context to take immediate action, underscoring its potential to compress red-team timelines, improve risk quantification, and close the gap highlighted by Mythos. Dror Kashti, Sweet Security’s CEO and co-founder, positions the product as a way for defenders to “debunk” the notion they cannot keep pace with AI-grade threats by safely executing those threats in their own environments before attackers do. Sweet Attack is now generally available to all Sweet Security customers, reinforcing the privately funded company’s strategy to anchor its cloud and AI security platform in real-time behavioral data and outcome-focused, board-ready insights.