AutoAlign Launches First Generative AI Sidecar to Bolster AI Safety

AutoAlign, an AI safety company that provides robust generative AI security, today announced the company’s spin off from Armilla AI, launching the first dynamic LLM security firewall solution Sidecar and cooperation agreement with professional services firm KPMG in Canada. The company’s new foundational Sidecar technology runs alongside LLMs — providing consistent security by dynamically interacting with, while continually evolving alongside, rapidly growing LLMs. In addition, KPMG will include AutoAlign as an option to its compliance layer for its Trusted AI framework. Now, any business using automation — including financial services — can tune Sidecar for their business compliance and security needs to deploy powerful, trustworthy generative AI.

“By the time companies build AI with the proper security guardrails, they’re immediately obsolete. That’s why we built our sophisticated AI firewall solution — Sidecar — so enterprises have LLM security that is comprehensive, nimble, and always evolving,” said AutoAlign interim CEO and co-founder Dan Adamson, whose two decades building AI solutions yielded products that are deployed at hundreds of enterprises today. “When generative AI exploded, it became clear that Armilla’s two business anchors — model security and AI insurance — required their own companies to provide services that customers deserved,” said Adamson, who will continue to stay involved with Armilla.

Faulty AI that hallucinates and allows jailbreaks, biased outcomes, and more is a byproduct of generative AI’s explosive growth as well as increased model complexities. That’s why AutoAlign created the first firewall security, Sidecar, that protects against these major shortcomings by interacting directly with LLMs. Sidecar’s adjacent rail structure with highly contextual Alignment Controls runs parallel to models, seamlessly moving from model to model and/or across use cases. By directly interfacing with LLMs, if Sidecar rejects a response, it will inform the model to make corrections. This cycle raises model fidelity by learning in real-time, and outcomes are meticulously honed on specific business use cases. Now, instead of fine-tuning a model to the point that it affects performance — or over-tuning its power away — enterprises can instead focus resources on increasing model performance.

“Working with KPMG exemplifies two enterprises committed to AI’s powerful and safe future,” said AutoAlign CTO and co-founder, Rahm Hafiz. “Providing customers with AI security solutions that they deserve is paramount at AutoAlign, no matter if AI is making internal decisions or used in customer-facing chatbots.”

“KPMG’s Global Trusted AI framework and approach to designing, deploying, as well as managing AI models and tools allow us to deliver powerful, trustworthy solutions to clients. AutoAlign shares our goal of integrating generative AI into businesses in a responsible manner without compromising the effectiveness of this transformative technology,” said Ven Adamov, Partner, National Leader of KPMG in Canada’s Risk Analytics practice and Trusted AI co-leader.\

Source:businesswire.com