The new capabilities in Ascend AI and Defend AI add automated red teaming, real-time guardrails and full-chain forensics to secure every agentic application
Straiker, an AI security company, unveiled new agentic security capabilities in its Ascend AI and Defend AI products to help enterprises confidently adopt AI agents. These new “attack and defense agents,” fine-tuned and trained on real-world agentic exploits, offer continuous security testing, automated enforcement, and chain of threat traceability, marking the industry’s first comprehensive solution for agentic AI threats.
Autonomous AI agents are replacing applications as we know them. According to the 2025 Stanford AI Index report, 78% of organizations were already using AI in 2024. With just an LLM and access to tools or datasets, agents can execute complex workflows in seconds. However, this power presents new risks, with attackers exploiting agents using natural language to create what Straiker calls autonomous chaos™. According to Straiker’s own research, 75% of tested applications were vulnerable to direct or indirect prompt injection attacks, which can lead to data exfiltration.
To address this challenge, Straiker offers full-spectrum protection, spanning prompt injection, reconnaissance, tool manipulation and exploit defense through enforcement and forensics. Straiker unifies security functions among AI, offensive testing and application security teams. Ascend AI provides autonomous agentic red teaming to craft highly accurate attacks and exploitation, while Defend AI instantly converts these never-before-seen attacks, like indirect prompt injection and tool vulnerability exploitation, into real-time guardrails that neutralize threats in production.
Straiker secures agentic AI applications with three key capabilities:
- Continuous Agentic Red Teaming: Ascend AI’s attack agents autonomously map every prompt, tool call, and data flow. They simulate realistic exploit chains to surface risks, integrating with CI/CD pipelines to deliver ongoing assessments and remediation guidance.
- Real-Time Agentic Guardrails: Defend AI automatically updates its guardrail engine in real time to block emerging threats targeting agentic AI applications, including tool misuse, vulnerability exploits, reconnaissance and excessive autonomy. It delivers protection instantly, with no code changes or added latency.
- Chain of Threat Forensics: When an attack occurs, Straiker reconstructs every prompt, decision and API call, using logs and sensor data to deliver a complete narrative. This traceability accelerates root-cause analysis, simplifies audits and demonstrates security value to stakeholders.
Aman Sirohi, CISO, People AI, said: “Straiker’s AI-native red teaming quickly adapted to our agentic AI application, enhancing our cybersecurity capabilities by providing guardrails to protect our AI agents from real-time exploitation and malicious behavior, thereby adding data security to our customer data.”
Also Read: Reality Defender & ActiveFence Partner to Bolster AI Safety Systems
Dan Garcia, CISO, EnterpriseDB, said: “Ascend AI stress-tested our entire agentic AI application stack, uncovering attack paths our manual red teaming exercises wouldn’t have been able to accomplish.”
CISO, FinTech: “We plugged Defend AI product in with a few lines of code and saw it apply guardrails across prompt injection, toxicity, PII leakage and other agentic threats in under a second, while showing us exactly where it happened. It’s the first solution that lets us push agentic features to production and sleep at night.”
Ankur Shah, co‑founder and CEO, Straiker, said: “If you can say it, you can spin up an autonomous AI agent and get it to perform tasks. That creative power deserves an equally autonomous defense. Straiker’s AI‑native security is built to learn, adapt and fight back in real time – so the future with AI stays safe.”
Source: PRNewswire
