Replacing the stop and think myth with conditioned reflexes at Signal Security Summit 2026

During internal red-teaming at SoSafe, our security team ran into an anomaly that should concern every CISO. Employees who routinely caught suspicious emails had almost no defence against malicious links handed to them by conversational AI. Decades of enterprise awareness training conditioned staff to scrutinise unexpected messages in an inbox, yet nobody has learned to distrust an interface that replies with articulate confidence.

It turns out scepticism is remarkably easy to disarm once software speaks politely.

At HuFiCon 2025—which returns to Cologne this November as the Signal Security Summit—SoSafe CISO Thomas Owen and CTO Rob Daly pointed out why the problem isn’t technical, but economic. Attackers operate with zero administrative overhead. While enterprise defenders complete data protection assessments and wait on procurement approvals, a solo attacker running autonomous agent fleets can scrape corporate context to deliver tailored lures for pennies. Their decision loop operates in seconds, unbound by governance.

When attacks arrive that fast with convincing phrasing, the traditional advice to stop and verify everything rings hollow. Employees cannot conduct manual forensic checks on inbound requests while keeping up with their daily workload.

Defence can no longer be an analytical task; it has to become an automatic behavioural reflex.

Deliberate analysis is a luxury busy people don’t have

Checking headers and inspecting domains lives strictly in deliberate, analytical thinking. Under heavy cognitive saturation, the human brain sidelines that deliberate analysis to conserve energy, relying on fast pattern-matching instead.

Crucially, social engineering does not succeed by catching careless lapses. It works by weaponising basic human decency, relying on our natural impulse to be collaborative and responsive to colleagues. When an urgent request lands during a busy workday, taking time to doubt a colleague feels uncomfortably close to insubordination.

Executive teams usually respond to emerging risks by tightening policy, but that only pushes the behaviour into the shadows. At HuFiCon 2025, Jasmine Jalava pointed out that banning technology works like telling a child they can never have sweets: you simply find wrappers hidden in their backpack. 

When a healthcare organisation restricted its staff to an approved enterprise Copilot, employees immediately turned to their personal phones to run prompts through external models. People were not looking to break rules; they were simply trying to deliver on commitments their managers were chasing them for.

The same dynamic undermines manual threat inspection. Jasmine highlighted how easily modern voice cloning incorporates subtle human flaws, like a hurried breath or background departure lounge noise. Human psychology is naturally suspicious of sterile perfection, so hearing natural conversational pauses immediately convinces the brain that the caller is genuine.

If security demands conscious deliberation while the clock is ticking, it will always arrive too late.

When attacks move in seconds, conscious thinking arrives too late

This structural mismatch is what Dr Niklas Hellemann refers to as the instinct gap: the distance between machine-speed attacks and the biological reality of human decision-making.

In his opening keynote at the Signal Security Summit 2026, The instinct gap: AI changed the rules. Now we need to adapt, Niklas reframes how organisations must approach human readiness. When synthetic media eliminates obvious tells, defence requires conditioned hesitation—an immediate gut check triggered by emotional pressure before rational analysis even begins.

In her session, Preparing employees for the age of AI-powered hyper-personalised attacks, Laura Adina explores how security teams can run realistic simulations across voice cloning and mobile channels, conditioning staff to recognise the physical adrenaline rush of executive urgency.

That shift also demands a complete reassessment of how we measure success. Click rates are a vanity metric easily skewed by arbitrary test difficulty; what truly matters is reporting velocity—the minutes between an employee suspecting an anomaly and raising the alarm. The single biggest barrier to reporting is the shame tax, the fear of looking foolish. When raising an alarm carries social embarrassment, that hesitation buys the attacker hours of uncontested dwell time.

Designing daily work so the right reflex happens by default

Reflexes only hold up if the surrounding environment supports them. In emergency search and rescue, as Thomas Owen saw firsthand, when a responder fumbles under adrenaline, you never blame the individual or issue a policy reminder. You examine the equipment layout, because muscle memory collapses the moment a medic has to hunt through a disorganised pack in the dark.

Cybersecurity remains one of the few disciplines that treats pressure-induced slips as personal disciplinary failures rather than flaws in system design.

Urban planners understand this through desire paths: if you erect a concrete fence across the natural route people take across the grass, they will simply climb over it. Enterprise security constantly builds concrete barriers across daily tasks, leaving employees no choice but to carve unmonitored dirt tracks on personal devices.

In their keynote, Adaptive learning, instinctive action: How to truly embed secure behaviour, Rob Daly and Christine Siu show how to pave those paths through adaptive micro-interventions tailored to specific roles and workflows, making the safe action the path of least resistance.

You don’t need another lecture explaining that phishing has evolved. What makes these sessions essential is the operational blueprint: conditioning physiological hesitation under pressure, removing the shame tax from incident reporting, and engineering workflows so the safe response is the easiest path.

On 18 November 2026, the Signal Security Summit brings over 2,000 security practitioners to Cologne to put those operational principles into practice.

PS: If you can’t join us in Cologne on 18 November, you can still follow the discussions and hear directly from the experts on stage. The Signal Security Summit 2026 will also be available via livestream.

Join the livestream