How Agentic AI Turns a £50k Fine Into £750k — Founder, Giovanni Coletta

A person makes one compliance mistake at a time. An agentic system makes the same mistake fifteen times in series, and a £50k fine quietly becomes £750k before anyone's read the logs. Giovanni Coletta on why good governance is far cheaper than the clean-up. Chapters: 0:00 Intelligence to AI governance: the throughline 0:34 The misspelt surname that broke a clean record 5:43 Why Reddit is the OSINT source analysts overlook 10:39 When the client ignores the risk you flag 14:54 The two governance failures behind most risk 21:02 Documented versus done: finding the gap 23:59 Governance by design, not patched in later 26:55 Can AI govern AI? 28:40 Quantum computing and the encryption risk ahead Giovanni Coletta is an AI governance professional and the founder of Quantidil. He spent the best part of a decade in corporate investigations and risk consulting at firms including Control Risks and Kraken, with Italy as a specialist focus. He now helps organisations let their staff use AI safely and responsibly. Two layoffs in three years, he says, sharpened how clearly he thinks about risk. Barnveil: https://barnveil.co.uk Host (LinkedIn):   / aarontheinvestigator   Guest org: Quantidal Giovanni Coletta spent close to a decade in corporate due diligence and risk consulting at Control Risks and Kraken before founding Quantidil, and that investigative background shapes how he reads AI risk. We get into the craft first: the Italian entrepreneurs who run at municipal elections under misspelt surnames, where a subject looks clean under their legal name yet carries a pile of adverse media under a variation buried in a public electoral list. Reddit comes up as the open source intelligence tool most analysts still sleep on. Not gospel, but a goldmine of leads, and the thing that once told Giovanni to run his research in a different language entirely. The OSINT mindset that maps a digital footprint, cross-references a registered address and treats data exposure as evidence is the same one he now applies to governance. His contrarian line: good governance is cheap. You don't need OneTrust and a fifteen-person team. You need someone who knows what good looks like, clear roles, escalation paths and a Power BI dashboard. Governance by design, before development, not bolted on after the fact. We close on quantum computing and the encryption risk almost nobody is pricing in yet. If you sit anywhere near intelligence, AI risk, compliance or privacy, this one earns your time. Follow Uninvited if you work anywhere near intelligence, AI risk or compliance. #OSINT #AIGovernance #Uninvited