Booking Holdings' security chief says generative AI has forced a fundamental rewrite of the travel industry's threat model over the past 18 months, expanding it from traditional infrastructure and application security into identity integrity, prompt injection, machine-to-machine trust and shadow AI adoption across the enterprise.

Devon Bryan, SVP and Global CSO at Booking Holdings — the parent of Booking.com, Priceline, Agoda and Kayak — laid out the shift in an interview with Help Net Security. His core argument: attacks that once demanded specialized skills can now be executed at speed, scale and sophistication that was previously unavailable, and travel's commercially interconnected ecosystem amplifies the damage.

Why travel is a compounding-risk business

Bryan, who moved from Air Force network security engineering through financial services, critical infrastructure and hospitality before joining travel technology, points to interconnectedness as the industry's defining security characteristic. Travel combines identity documents, payments, loyalty point economies, third-party integrations, global operations and geopolitical exposure at enormous scale.

"A disruption in one area can create downstream operational, financial, and reputational impacts across customers, partners, and employees," he said. Attackers, he added, increasingly target trust relationships and operational dependencies rather than individual systems — meaning the attack surface now includes vendors, APIs, cloud environments, partner ecosystems and AI-enabled workflows.

The revenue implication for travel sellers is direct. Loyalty point economies and stored payment credentials are monetizable assets for fraudsters, and a breach at one integration point propagates across a distribution chain that connects OTAs, suppliers and payment rails. For a business processing millions of transactions, Bryan argues resilience — detection, response and recovery — matters as much as prevention.

What changed in 18 months

Bryan is specific about how AI has reshaped the threat landscape. Phishing, impersonation, fraud and social engineering have become more personalized, multilingual and operationally scalable. His threat modeling now extends to AI-generated content, model access controls, data lineage, third-party AI dependencies and unsanctioned employee use of AI tools.

At the same time, he positions AI as a defensive asset rather than only a threat. Booking Holdings is using it to improve threat detection, vulnerability prioritization, fraud analytics and operational efficiency across security workflows. "AI is simultaneously a force multiplier for both attackers and defenders, which means organizations must approach it with both optimism and discipline," he said.

That dual framing carries cost and governance consequences for any travel platform embedding AI into booking flows, customer service automation or dynamic pricing. Securing model access and monitoring shadow AI adoption are now line items that did not exist in most travel security budgets two years ago.

When security gets a vote

Bryan described a four-part filter for deciding when security engages in business decisions — M&A diligence, product strategy, AI adoption, regulatory discussions, geopolitical risk: trust, resilience, regulatory exposure and systemic business impact. If a decision could materially affect customer trust, operational continuity or the ability to scale securely, security needs a voice early.

But he draws a hard line against security operating as "a perpetual veto function." In large federated organizations, he said, the goal is not centralized control over every decision but clear standards, accountability models, escalation paths and guardrails that let teams move quickly within an acceptable risk framework.

"Maturity in security leadership is often about knowing where not to over-rotate," Bryan said. "If security inserts itself too deeply into every operational decision, you create friction and dependency. The best security organizations enable confident decision-making."

The hiring signal: judgement over technical depth

Asked what skill is most undervalued in how the industry hires and develops security leaders, Bryan's answer is judgement — the ability to convert technical signals into sound business decisions under pressure and ambiguity. He has watched technically brilliant people struggle because they treated every issue as a technical problem, missing context on timing, customer impact, legal considerations and risk tolerance.

Judgement also shows up in communication: simplifying complexity for engineering, legal, finance, operations and the boardroom. "That is increasingly what separates strong operators from future enterprise leaders," he said.

Looking ahead, Bryan frames the next generation's mandate as managing AI, geopolitical instability, regulatory expansion, supply chain complexity and increasingly autonomous systems simultaneously — a signal that travel's largest distribution groups intend to treat security leadership as an enterprise function, not a technical cost center, as those pressures compound.