A hands-on session where participants put AI tools under pressure and watch how human behaviour shapes what comes back.

Working in small groups, participants use everyday AI assistants on realistic security scenarios: assessing a suspicious message, reviewing an AI-generated recommendation, and deciding what to trust. Each round surfaces a different human vulnerability: how framing steers the output, how a confident first answer anchors every answer after it, and how social engineers exploit the same shortcuts in people that AI reproduces at scale.

We then rebuild each scenario with simple behavioural checks: stating priors before the prompt, testing for confirmation bias after the first answer, and building human-review points that people will actually use. The session draws on the AI and Digital Security Interaction Workshop delivered at LSE in August 2026 (52 participants, 95% met their goals) and on MindHug's CPD-accredited digital-security training for DSIT/Innovate UK Cyber Local.

Participants leave with a reusable checklist for AI-assisted security decisions and a clearer view of where their own team's blind spots are.

Bring a laptop with access to any mainstream AI assistant (ChatGPT, Copilot, Claude or Gemini).