HiddenLayer raises $50M to defend enterprise AI models

HiddenLayer, an Austin, Texas-based cybersecurity startup born out of a cyberattack that exploited machine studying code on the founders’ prior firm, has introduced a $50 million Collection A funding spherical at this time to additional harden the defenses of the rapidly growing number of AI models being adopted by enterprises.

The spherical was led by M12, Microsoft’s Venture Fund, and Moore Strategic Ventures, with participation from Booz Allen Ventures, IBM Ventures, Capital One Ventures, and Ten Eleven Ventures.

“AI’s unapparelled price of adoption fuels us to maneuver even quicker in reaching our mission to provide each safety skilled the proper instruments and experience for embracing AI securely,” mentioned Chris Sestito, CEO and Co-Founder at HiddenLayer, in a press release within the firm’s press launch asserting the spherical.

Already, HiddenLayer helps safeguard AI/ML fashions utilized by a lot of Fortune 100 corporations throughout sectors inclucing finance, authorities and protection, and cybersecurity.

What HiddenLayer does

As beforehand lined by VentureBeat final yr following its emergence from stealth, HiddenLayer has constructed a lot of instruments as a part of its “MLSec” Platform for safeguarding enterprise machine studying (ML) and AI fashions.

These instruments don’t really entry the fashions, nor compromise the proprietary knowledge and know-how of purchasers. As a substitute, the passively monitor the efficiency and operations of enterprises ML/AI fashions and linked functions in realtime, scanning overarching vulnerabilities and providing suggestions for hardening them, in addition to detecting injection of malicious code/malware and deploying protection mechanisms to chop off the attackers and isolate any intrusions.

HiddenLayer’s MLSec Platform ships with a easy however highly effective dashboard permitting safety managers to get entry to all the knowledge they want in regards to the safety state of their enterprise ML/AI fashions at a look. It additionally routinely lists safety points and alerts so as of precedence relying on the severity of the problem, and shops knowledge for the compliance, auditing and reporting {that a} enterprise could also be requested to do.

HiddenLayer additional presents consulting companies from its staff of Adversarial Machine Studying (AML) specialists who keep atop the most recent tendencies in safety and the most recent threats. They’ll carry out risk assessments, coaching for a consumer’s cybersecurity and dev ops personnel, and carry out “red team” workout routines to make sure the consumer’s defenses are working as meant.

Influential accomplice

Earlier this yr, the corporate struck a partnership with white-hot enterprise knowledge lakehouse supplier Databricks, permitting Databricks enterprise prospects to make use of HiddenLayer’s MLSec Platform straight on their fashions operating on Databricks’ lakehouses.

“The mixing is mannequin agnostic and contains mannequin scanning and mannequin detection and response,” defined HiddenLayer on the time in a blog post announcing the partnership. “This allows Information Scientists and ML Engineers so as to add safety to their fashions with no code or behavioral modifications to their setting. Because the mannequin is loaded, will probably be scanned by HiddenLayer’s mannequin scanner to make sure integrity in addition to safety. If an assault is detected, the combination will deal with the response accordingly with none human interplay wanted.”

What’s subsequent for HiddenLayer’s quest to safe enterprise AI?

HiddenLayer was based after co-founders Sestito (CEO), Tanner Burns (chief scientist) and Jim Ballard (chief info officer) after the three encountered a cyberattack on ML fashions on the prior firm, Cylance, a safety startup.

As recollected on HiddenLayer’s website, the incident occurred when “attackers had exploited Cylance’s Home windows executable ML mannequin utilizing an inference assault, exposing its weaknesses and permitting them to supply binary recordsdata that might efficiently evade detection and infect each Cylance buyer.”

Whereas worrisome and irritating on the time, the trio realized then that assaults on ML/AI would solely enhance within the close to future as extra enterprises sought to undertake generative AI into their workflows as a result of know-how’s nice promise at growing effectivity and efficiency.

As we speak, HiddenLayer is rising quickly, having quadrupled its headcount within the final yr. Now flush with its Collection A money, it plans to rent one other 40 personnel by the yr’s finish, and cont proceed rising its consumer base.