CalypsoAI Unveils Autonomous AI Agents as Primary Cyber Threats

CalypsoAI has announced that autonomous AI agents have become the primary drivers of cyber threats, surpassing human hackers in threat research, intelligence gathering, and attack generation announced in a press release. The August CalypsoAI Security Index (CASI) Leaderboard revealed a 12.5% drop in security scores across all major AI models, highlighting the significant impact of these agentic systems.

The company's Inference Red-Team product, powered by Signature Attack Packs, enables fully autonomous agents to research, generate, test, and curate high-severity adversarial prompts. This capability allows for the exposure of real-world vulnerabilities at a speed and scale unmatched by human teams.

CalypsoAI's enhanced Inference Red-Team solution now includes Agentic Fingerprints, providing customers with detailed insights into attack agents' behaviors, decision-making processes, and successful exploits. Additionally, the Inference Defend solution has been upgraded with Outcome Analysis, offering clear visibility into why certain prompts and responses are flagged or blocked.

CalypsoAI introduced these upgraded solutions at Black Hat USA 2025, emphasizing the importance of understanding AI systems' vulnerabilities and failures as enterprises increasingly adopt AI applications.