The Competitive Edge The Secret Agent: Peter Waters and Andrew Low on the rise and the risks of agentic AI in commerce
Apr 1, 2026
Andrew Low, a competition and consumer regulation lawyer, explains legal risks around contracting and liability for agentic AI. Peter Waters, telecoms and AI policy commentator, maps how AI agents plan, act and sometimes misbehave. They discuss agentic AIs organising real-world transactions, who bears risk when agents overstep, and what regulatory building blocks could keep consumers safe.
AI Snips
Chapters
Transcript
Episode notes
How Agentic AI Steps Out Of The Screen
- Agentic AI augments an LLM with memory, internet input tools, output/payment tools, and a planning module so it can act and transact autonomously.
- Peter Waters' ski-trip example shows it can scan emails, set preferences, book flights/hotels, and pay with your card.
First Real Agentic Transaction In Australia Example
- Mastercard ran an Australian transaction where an AI agent autonomously used a Westpac card to book ski accommodation.
- Peter Waters used this example to show agentic AI is already transacting with real payment systems.
Existing Case Law Treats AI As A Legal Tool
- Courts so far treat AI as a tool: when algorithms enter contracts mistakenly, principals remained liable because the algorithm acted per its programmed parameters.
- Andrew Low cites Singapore and Canadian cases where deals were enforced against the human operator despite algorithmic errors.