← Speaking & Keynotes
AI-Enabled Deception

Don’t Train Humans
to Spot the Fake

Build systems that remain safe when synthetic content becomes indistinguishable from authentic content.
Discuss this keynote
Human vigilance cannot be the final security control.

The current defense model has a structural weakness.

Organizations routinely train people to identify suspicious grammar, unusual voices, visual artifacts and inconsistent behavior. Those defenses matter today — but they depend on deception continuing to reveal itself.

As synthetic media improves, the long-term strategy cannot be to ask humans to become increasingly sophisticated authenticity detectors.

Traditional Model
Content arrives
Human examines it
Human decides whether it looks fake
Human acts
Resilient Model
Request arrives
Independent verification
Context check
Appropriate friction
Trusted confirmation
Action
“The request cannot authenticate itself.”

Verification must come from outside the interaction.

A caller claiming to be an executive cannot establish that claim solely with evidence provided during the same call. A message cannot become trustworthy simply because it contains another link. A synthetic video cannot establish its own authenticity merely by appearing convincing.

Trust needs an independent source of confirmation.

Verify Confirm through an independent trusted mechanism.
Context Ask whether the request makes sense in the surrounding circumstances.
Friction Slow consequential actions where additional certainty has value.
Network Use trusted people and channels outside the interaction itself.
“The goal is not perfect fake detection. The goal is safe decision-making even when fake detection fails.”

Design beyond human perception.

A keynote for security, fraud, technology, trust & safety and leadership audiences confronting increasingly convincing synthetic deception.

Discuss this session