The same AI tools that pharmaceutical companies use to accelerate drug discovery are being used by criminal networks to engineer new psychoactive substances that don’t yet exist on any banned substances list. That’s the warning from Dr Lorraine Nolan, Executive Director of the EU Drugs Agency (EUDA), speaking to the Financial Times on 9 June 2026 ahead of the European Drug Report 2026.
“It definitely can be utilised in the realm of designer chemistry and the production of illicit drugs,” Nolan said. “When you examine the technological advancements in the pharmaceutical sector and how they are accelerating drug discovery, the same principles apply to the illicit side.”
The process is simple: AI can forecast molecular structures with specific pharmacological effects – legitimate researchers use this to find new medicines faster. Criminal organisations are using the same capability to design “designer precursors” – close chemical relatives of controlled substances, purpose-engineered to produce similar effects while sitting just outside existing legal definitions.
Approximately one new psychoactive substance was identified per week in Europe last year, with particularly potent synthetic opioids among them.
The Problem Regulators Can’t Easily Solve
This isn’t a new issue, but AI has definitely made it a lot tougher to manage. Drug control law operates on a list-based system: a substance is either scheduled (controlled) or it isn’t. Criminal producers have always exploited that space by tweaking chemical compositions just enough to produce an unscheduled variant. What AI changes is the speed and sophistication at which new variants can be identified and produced.
The UN International Narcotics Control Board made the same point in 2024, warning that criminal groups are exploiting AI to find alternative chemicals for drug production. The volume of synthetic substance seizures has surpassed that of heroin and cocaine. Synthetic drugs can be manufactured anywhere without large-scale cultivation, making them cheaper and harder to trace than traditional narcotics. Europe, according to EUDA, is developing into a hub for this kind of synthetic production.
In April 2026, the European Commission brought nine high-risk precursor chemicals under regulatory control via Delegated Regulation (EU) 2026/314, covering eight substances linked to synthetic cathinones and one linked to amphetamine production, applying from September 2026. Revised precursor regulations introduced in 2025 also created faster mechanisms for adding designer precursors to the controlled list. EUDA has upgraded its early warning system to alert laboratories, law enforcement and health services about new active substances.
The problem is the speed gap – regulators identify a substance, assess it, consult, draft legislation, pass it and enforce it – that process takes months to years. An AI model generating new molecular candidates operates in seconds. By the time a precursor is scheduled, a structural variant is already in circulation.
More from Artificial Intelligence
- Google’s AI Price Cuts Suggest Subscription Pricing Could Define The Next Phase Of AI Competition
- AI Has Moved Past Testing In Most Businesses – So What Comes Next?
- 87% Of Brits Know Little About Data Centres Despite Using Them Every Day
- This Fund Is Investing In Researchers Before They Even Start A Company
- Project Glasswing: How Will It Impact The Cybersecurity And AI Industries?
- Are We Too Reliant On Anthropic’s Claude?
- AI Is Now Better At Catching Fraudsters Than Human Analysts: Here’s Why That Matters
- World Environment Day 2026: How Tech Is Impacting Europe’s Climate Footprint
The Dual-Use Reality Nobody In Tech Talks About Enough
The AI capabilities driving this problem are the same ones being celebrated in legitimate biotech and pharmaceutical research. There’s no technical distinction between an AI model that helps a pharmaceutical company discover a new painkiller faster and one that helps a criminal network identify a new synthetic opioid that evades the current banned substances list. The tool is the same, the intent is different and the outcomes diverge enormously.
EUDA notes that AI could also support anti-drug strategies through earlier threat detection – analysing patterns in seizure data, flagging emerging compounds before they reach the street, modelling supply chain disruptions. The technology cuts both ways, but the question is whether law enforcement and regulatory agencies can use it as effectively as the networks they’re trying to stop, and whether the institutional speed of government can close the distance that AI has opened.
The EUDA warning is a reminder that dual-use risk is a live regulatory and reputational issue. Organisations in the chemistry and life sciences space who are thinking hardest about it right now will be best positioned when the policy response catches up with the technology.
