Novee Launches AI Red Teaming Platform to Stress-Test LLM Applications for Security Flaws

New updates have been reported about Novee.

Novee has launched an AI red teaming capability for large language model (LLM) applications, positioning its AI penetration testing platform to target security gaps in rapidly deployed AI-enabled software such as chatbots, copilots, autonomous agents, and workflow tools. The platform’s autonomous agent continuously simulates advanced, real-world attack patterns—including prompt injection, jailbreaks, data exfiltration, and agent manipulation—that traditional web and infrastructure testing tools typically fail to detect.

The system evaluates how LLM-powered applications respond under adversarial conditions, then generates vulnerability assessments and concrete remediation steps that can integrate into existing security testing workflows and CI/CD pipelines. Novee is training the agent on techniques distilled from its own offensive security research, including a recently disclosed vulnerability in Cursor that enabled attackers to hijack a coding agent’s context window and execute remote code on a developer’s machine.

CEO and co-founder Ido Geffen emphasized that attackers are compressing the time between discovery and exploitation to minutes, arguing that this shift requires continuous rather than periodic testing for AI systems. Co-founder and CPO Gon Chalamish added that most organizations still test AI applications with legacy web-focused tools, while adversaries are already evolving tactics tailored to LLMs and autonomous agents.

Novee’s agent is architected to work across any LLM-based application regardless of underlying model provider or stack, supporting deployments on commercial platforms like OpenAI and Anthropic as well as open-source models. This model-agnostic approach is intended to make the tool relevant for large enterprises standardizing AI development while seeking to unify security coverage across heterogeneous AI infrastructures.

The company is initially releasing the AI red teaming product in beta and is using RSA Conference 2026 as a key go-to-market milestone, including live demonstrations to security buyers. Findings from active research and ongoing responsible disclosures to multiple vendors are being fed back into the agent’s training loop, aiming to keep Novee’s detection logic aligned with emerging attacker techniques and to strengthen its positioning in the fast-growing AI security segment.

Novee, founded by national-level offensive security veterans Ido Geffen, Gon Chalamish, and Omer Ninburg, has raised $51.5 million within four months of its inception from investors including YL Ventures, Canaan Partners, and Zeev Ventures. This funding base provides runway to scale product development and commercial deployment of its AI penetration testing platform as enterprises accelerate adoption of AI-driven applications and face mounting regulatory and operational pressure to secure those systems.