Withinai announces automated redness and next-generation AI firewall protection for enterprise LLMS and AI applications

New AI security feature tests and enhancements to enhance LLM against real-world threats while enforcing security and governance policies during live use

Following a quarter of record sales, Withinai, the Secure AI Enablement Leader, has announced two new products that protect enterprise LMS and AI applications through automated red teams and behavioral runtime protection. Together, they provide global companies with automated features for testing, hardening and defense, enterprise AI models, and apps and agents that use them.

• Witness attack An automated red teaming tool designed to actively identify vulnerabilities in large-scale language models (LLMS) before deployment. Generate synthetic prompts that emphasize the defense of your model using simulated attack techniques such as multi-modal attacks, multi-step jailbreak, comprehensive fuzzing, API integration, and augmented learning attacks. Eyewitness attacks allow developers to easily identify and address weaknesses in their own defenses before they are deployed.
• Witness protection is a next-generation AI firewall that provides action runtime defense against model attacks such as rapid injection and jailbreak. It also provides LLM defenses such as behavioral prompt filtering, intention-based response control, real-time data editing, threat detection, harmful responses, toxicity filtering, and model ID enforcement. Witness protection includes two years of synthetic rapid data and new defence techniques honed from continuing conversational attacks. As a result, this solution offers industry-leading detection of model attacks and has more than 99% effective identification of rapid injection. Witness Protection Defense provides standardized protection across models and works with over 100 LLMs.

Also Read: Aithority Interview with Suzanne Livingston, Vice President of IBM Watsonx Orchestrated Agent Domain

With the introduction of witness attacks and witness protection, the Withinai platform addresses a wide range of enterprise AI security and compliance requirements, including secure AI development, employee AI use, secure models and app runtime operations.

“Companies don't want to buy five different products to ensure employees and customers can use AI safely,” he said. Rick KucchiaWithinai CEO. “With the introduction of witness attacks, businesses can now ensure automated testing and hardening of internally developed models, apps and agents. Witness Protection adds even better defense against model attacks and is already receiving customer ratings.

Also Read: c-gen.ai impacts AI workloads from stealth to end-infrastructure limits