AutoAligna number one provider of AI security solutions, announced Tuesday that it has spun off from its parent company vest to concentrate on protecting and improving AI systems for enterprise customers. The company also introduced its flagship product: “sidecar“, a brand new approach to AI security that works with existing models to detect and mitigate potential problems reminiscent of data hallucinations, jailbreak attempts, and bias.
The spin-off comes as Armilla shifts its focus to providing AI systems risk transfer solutions to the insurance industry. “There's actually an inherent conflict between protecting (AI systems) and making them more robust on the one hand after which insuring them on the opposite,” said Dan Adamson, interim CEO and co-founder of AutoAlign, in an interview with VentureBeat. “Both parts of this are gangbusters… nevertheless it's really necessary that they're done as separate entities.”
Addressing the growing need for AI security
As more corporations look to adopt AI technologies, the necessity for robust security solutions is becoming increasingly clear. “We see loads of prototyping. However, due to these issues, we don’t see lots of them going into production,” Adamson said. “It's an enormous problem… over three-quarters of (CIOs and CTOs) have as their primary concern the risks related to adopting AI systems because the predominant reason for not bringing them into production.”
AutoAlign's Sidecar platform goals to deal with these concerns by providing a contextual, adaptive approach to AI safety. “You can maintain or preserve the performance of those more powerful models while making them protected and effective,” Adamson told VentureBeat.
Operationalizing responsible AI with KPMG partnership
AutoAlign also announced a collaboration agreement with KPMG Canada to map the consulting giant's trusted AI policy framework to the actual alignment controls within the AutoAlign system. “KPMG has been a frontrunner in responsible AI for a while and has a trusted AI framework that is essentially on the policy level,” Adamson said. “And what we've achieved with this agreement is that we're now in a position to map those policies to actual alignment controls inside Autoalign, so that they can map any principle to those controls.”
The goal of this partnership is to assist KPMG Canada clients implement responsible AI practices and speed up the deployment of AI systems in production. “For their customers, they’ll assign these controls once after which keep them as a sidecar. And as their models evolve, or as they reuse or construct on different use cases, they’ll keep and preserve the sidecar that meets those guidelines,” Adamson said.
Spinout positions AutoAlign for focus
During our interview, Adamson also demonstrated how the Sidecar platform can force models to supply factual, referenced answers on sensitive topics moderately than refusing to have interaction, a typical limitation of other AI security approaches. “You can imagine other use cases where we’ve got technical documentation on sensitive topics like medical devices, where the models are like, 'I don't feel comfortable answering on medical devices,'” he explained.
This capability could prove crucial for corporations trying to deploy AI in areas reminiscent of healthcare, finance and politics, where accuracy and context are paramount. By enabling models to supply nuanced, contextual responses, AutoAlign's Sidecar platform could help corporations address the complex ethical and safety challenges related to adopting AI.
We are able to impact the enterprise AI security market
With the support of Armilla and an existing partnership with KPMG, AutoAlign is well positioned to make a major impact on the enterprise AI security market. The company's unique approach, which focuses on improving moderately than limiting AI models, could prove to be a key differentiator in a rapidly evolving landscape.
As the capabilities of base models advance at an unprecedented pace, solutions like Sidecar are prone to grow to be increasingly necessary for corporations in search of to harness the ability of AI while ensuring the responsible and protected use of those systems. The launch of AutoAlign represents a vital milestone in the event of AI security technologies, paving the best way for greater adoption and innovation across industries.
