AI agent security
AI agents now act, not just answer. I research controls that limit the harm an agent can do, including an agent that is compromised or trying to get around its limits.
AI security · Information systems · Human behaviour
Keeping AI agents, and the people who rely on them, safe from harmful actions and deception.
I'm Dr Brent Coker, an information systems researcher at the University of Melbourne. I study where system design meets human behaviour, and I now focus that work on AI security: controlling what AI agents are allowed to do when the stakes are high and the agent may not be trustworthy.
Focus areas
AI-enabled deception
My background is in how people are persuaded and misled online. AI makes deception cheaper and more convincing, and it also makes AI systems themselves targets for manipulation.
AI governance and evidence
As an expert witness I have given evidence on fraud, privacy and data manipulation. AI incidents are reaching courts and regulators, and they need the same careful, evidence-based analysis.
Research
In progress
INTERLOCK: budgeted control of high-consequence AI actions
How do you stop an AI agent from taking an action that puts critical infrastructure or public safety at risk, even if the agent is working against you? INTERLOCK treats an agent's ability to cause harm as a budget that is spent as it acts, with hard caps on damage that can't be undone and controls that hold across organisations.
The work draws on a corpus of more than 3,700 papers on AI control and security. Papers and a plain-English explainer will appear here as they are published.
About
I hold a PhD in electronic commerce and information systems from Victoria University of Wellington, and I have been a Senior Lecturer at the University of Melbourne since 2008, teaching digital marketing and digital business models from undergraduate to MBA level.
My research on consumer behaviour and online decision-making has been published in the Journal of Experimental Social Psychology, the Journal of Retailing and Harvard Business Review. I have been appointed as an expert witness in matters involving consumer deception, fraud, privacy and data manipulation.
I also build. I founded a dot-com in 1998, invented the web analytics tool Webreep, and today I direct Deloosh, where I build and security-review AI systems in production.
Selected publications
- Coker, B. (2024). Make your marketing a force for good. Harvard Business Review, September.
- Coker, B. (2024). Beyond the buzz: Viral marketing under a legal lens. Australian Intellectual Property Law Bulletin, 36(5/6).
- Coker, B. & McGill, A. L. (2020). Arousal increases self-disclosure. Journal of Experimental Social Psychology.
- Coker, B. (2020). Arousal enhances herding tendencies when decision making. Journal of Consumer Behaviour, 19(3), 229–239.
- Coker, B. & Nagpal, A. (2013). Building-up versus paring-down: Consumer responses to recommendations when customizing. Journal of Retailing, 89(2), 190–206.
- Coker, B. (2012). Seeking the opinions of others online: Evidence of evaluation overshoot. Journal of Economic Psychology, 33, 1033–1042.
- Coker, B., Hope, B. & Ashill, N. (2011). Measuring internet product purchase risk. European Journal of Marketing, 45(7/8), 1130–1151.
- Coker, B. (2016). Going Viral. London: Pearson.