SBN

Guardrails, Not Gates | Eric Galis on Securing AI Without Slowing the Business

In Episode 3 of The Cyber Roundtable: Security Evolutions, I sat down with Eric Galis, Cybersecurity Advisor and Fractional CISO at Plaintext Security Advisors.

Eric has spent more than 20 years in security. He started at PwC during Boston’s early security boom, working with regional banks and credit card processors where the job ran on the stick of regulation and clear-cut fines. Then he moved into EdTech, where there was no stick to point to, and he had to lead through influence instead, protecting a product that kept changing shape, from a CD in the back of a textbook to a full interactive platform.

The thread that ran through our whole conversation: security works better as a guardrail than a gate. Eric pointed to Mario Kart’s Rainbow Road, a track with no edges and no rails, where you spend the whole race falling into space. Good security, he argues, doesn’t stop people at a checkpoint. It keeps them on the road and nudges them back when they drift, so the business can keep moving at speed.

Eric breaks AI security into three questions

  1. How do you secure against AI?
  2. How do you secure with AI?
  3. How do you secure the use of AI?

Most of what falls into the first bucket isn’t new; it’s old attacks running faster and on a larger scale. The genuinely novel problems, like deepfakes, live in integrity: knowing that what you’re interacting with is what it claims to be.

Some of the most practical ground we covered was AI clearing away toil. Eric described pointing a natural-language model at a company’s own policies, so a developer can ask, “Can I put secrets in GitHub?” and get an answer in the moment. Or feeding a customer’s security questionnaire and your existing documentation into a model that drafts 85 to 90% of the responses, handing a team back hours of its week.

And he named a risk most teams are quietly working around: the AI “hot potato.” Legal owns copyright, security owns the model, but the net-new risks AI introduces often have no owner at all. Eric’s view is that finding and claiming those gray areas is how a security leader shows they’re driving the business forward, not just guarding it.

A thoughtful conversation about moving fast without falling off the track. Take a listen.

Connect with our guest

Eric Galis

Cybersecurity Advisor & Fractional CISO, Plaintext Security Advisors

Visit Plaintext Security Advisors →

The post Guardrails, Not Gates | Eric Galis on Securing AI Without Slowing the Business appeared first on Realm.Security.

*** This is a Security Bloggers Network syndicated blog from Realm.Security authored by John Greene. Read the original post at: https://realm.security/eric-galis-the-cyber-round-table-episode-3/