Agents After Dark

Sara Abak on securing the enterprise while AI outruns the rules

Sara Abak asks a simple question: would a carmaker ship a vehicle without a seatbelt?

In this episode

  • Regulate frontier AI like any product that can cause harm at scale
  • Anchor security in the business, because you cannot protect what you cannot see
  • Board trust beats laying down the law and the department of no
  • Fund AI work inside existing projects with legal, HR, and IT
Full transcript click to expand

Transcript coming soon.

Frequently asked questions

What is Sara Abak's core argument about frontier AI?

She compares it to shipping a car without a seatbelt. Powerful AI models release with little regulation, and she argues unsafe releases should be paused rather than innovation blocked.

How does she approach turning around a security culture?

She anchors the work in what the business actually does, on the principle that you cannot protect what you cannot see. From there she builds trust at the board level rather than laying down rules.

How does she handle AI features in existing tools?

AI features in an ERP or board paper platform stay switched off until they pass a risk assessment. She treats each new capability as something to evaluate before enabling.

What is her approach to shadow AI?

She locks down what staff can paste into public AI tools, coaches people at the point of upload, and uses quarterly behaviour reporting so decisions are anchored in data rather than assumptions.

Why do standalone responsible AI programs struggle?

They rarely get funded on their own. She anchors AI work inside an existing project with legal, HR, and IT at the table, which is what actually moves things forward.

Stay ahead of the curve

No spam. Unsubscribe anytime. A resource by Prefactor.

Almost there — check your inbox to confirm your subscription.