By
Stuart Kerr, Technology Correspondent, LiveAIWire

The dark web marketplace AlphaBay processed over a billion dollars
of transactions before it was shut down by law enforcement in 2017. Its
successor markets are more sophisticated, more resilient, and increasingly
AI-assisted. Vendors use AI to write product descriptions that evade
automated scanning. Buyers use AI to identify quality vendors from coded
reviews. Administrators use AI to detect and exclude law enforcement
infiltrators based on behavioural patterns. The shadow economy has adopted
the same tools as the mainstream economy, and in some respects it has adopted
them faster.

AI is transforming organised crime, financial fraud, and the
broader shadow economy in ways that challenge law enforcement agencies whose
capabilities and legal frameworks were designed for an earlier technological
environment. The same capabilities that make AI commercially powerful —
natural language generation, pattern recognition, automation at scale — make
it operationally powerful for criminal actors who face no regulatory
constraints on how they deploy it.

AI-Enabled Financial Crime

Financial fraud has been transformed by AI at both the attack and
defence layers. On the attack side, AI enables the generation of synthetic
identities, the automation of account takeover attempts at scale, the
creation of convincing phishing communications personalised to individual
targets, and the development of deepfake voice and video used in authorised
push payment fraud — where victims are deceived into transferring funds to
accounts controlled by criminals.

Authorised push payment fraud losses in the UK reached nearly half
a billion pounds in the first half of 2024 according to UK Finance data. The
crime typically involves a convincing impersonation of a bank, employer, or
family member delivered via phone call or video — an impersonation that has
been dramatically improved by AI voice and video synthesis. The victim is not
hacked; they are deceived. The AI does not break the security system; it
bypasses it through the human in the loop.

Research from the Financial
Action Task Force on AI and financial crime has documented the use
of AI tools by organised criminal groups across multiple jurisdictions,
noting that the adoption of AI by criminal actors is accelerating in parallel
with legitimate commercial adoption and that the regulatory response has not
kept pace with the operational reality.

AI in Drug and Contraband Markets

Dark web drug markets have evolved into sophisticated commercial
operations that use AI for multiple operational functions. Automated vendor reputation
scoring, AI-generated product listings optimised for search visibility,
customer service chatbots that handle order inquiries, and AI-assisted
logistics optimisation for postal drug distribution are all documented in law
enforcement intelligence assessments of contemporary dark web
markets.

The operational security benefits of AI for criminal actors are
significant. Tasks that previously required human involvement — writing
product descriptions, responding to customer queries, screening potential
buyers — can be automated, reducing the number of people involved in
operations and therefore the number of potential informants. The fewer human
touchpoints in a criminal operation, the harder it is for law enforcement to
infiltrate it through conventional human intelligence
methods.

The connection to the broader challenge of AI
in law enforcement and criminal justice is direct: the same
algorithmic tools being deployed to detect criminal patterns are being
deployed by criminals to evade detection. The adversarial dynamic is
symmetric in a way that the arms race metaphor, with its implication of one
side outrunning the other, does not fully capture.

Human Trafficking and AI Exploitation

Human trafficking networks have adopted AI tools for victim
recruitment, particularly in online grooming and fraudulent employment
advertising. AI-generated job advertisements that target economically
vulnerable individuals, AI-assisted communication that maintains multiple
recruitment conversations simultaneously, and deepfake-enhanced social media
profiles used in romance fraud that leads to trafficking are all documented
in reports from anti-trafficking organisations.

The International Justice Mission and other anti-trafficking
organisations have developed AI tools to counter these tactics: scanning
platforms for suspicious advertising patterns, detecting AI-generated profile
images in exploitation contexts, and identifying the linguistic signatures of
AI-assisted grooming communications. This counter-application of AI against
AI-enabled trafficking represents a genuine use case for the technology in
harm reduction, though the scale of deployment by criminal actors currently
exceeds the counter-deployment by defenders.

Law Enforcement AI and Its Limits

Law enforcement agencies in the US, UK, Europol, and Interpol have
invested significantly in AI tools for financial crime detection, dark web
monitoring, and criminal network analysis. AI systems that identify
transaction patterns consistent with money laundering, that scan dark web
marketplaces for emerging criminal activity, and that map relationships
between individuals in criminal networks from communications metadata are all
in operational use.

The legal constraints on law enforcement AI vary significantly by
jurisdiction and create genuine friction with the technical capabilities
available. Evidence obtained through AI analysis must meet admissibility
standards that were not designed with algorithmic evidence in mind. Privacy
protections limit the data that law enforcement can access for AI analysis.
International coordination challenges mean that criminal operations that span
jurisdictions can exploit the gaps between different national legal
frameworks.

The Europol
Innovation Lab report on AI and the criminal world has documented
both the criminal adoption of AI and the law enforcement response, concluding
that the balance of advantage varies by crime type and that sustained
investment in law enforcement AI capabilities and international coordination
is required to prevent criminal actors from maintaining a persistent
capability advantage.

Regulatory Responses and the Encryption Debate

The policy response to AI-enabled crime intersects with
longstanding debates about encryption, platform regulation, and surveillance
powers. Law enforcement agencies consistently argue that strong encryption
and end-to-end messaging platforms impede access to evidence of serious crime
including terrorism, child exploitation, and organised crime. Technology
companies and civil libertarians argue that weakening encryption to enable
law enforcement access would compromise the security of billions of
legitimate users.

AI adds complexity to this debate in both directions. AI can
analyse metadata and behavioural patterns to identify criminal activity
without accessing encrypted content, potentially reducing the need for
decryption access. AI can also be used by criminal actors to more effectively
exploit end-to-end encrypted platforms for operational communications.
Neither consideration resolves the fundamental tension, but both are relevant
to how the policy debate develops in an AI-integrated communications
environment.

The broader
question of privacy rights in an AI-surveillance environment frames
the law enforcement debate: the capabilities that would make law enforcement
most effective against AI-enabled crime are the same capabilities that would
most significantly erode the privacy of the population that is not engaged in
crime. Where the balance should be drawn is a democratic question that
technical capability alone cannot answer.

The governance of AI-enabled crime ultimately requires
international coordination that the fragmented current landscape of national
law enforcement and legal frameworks does not support. Criminal actors that
exploit jurisdictional gaps, operate across borders, and adopt AI tools
faster than regulatory frameworks can respond will continue to outmanoeuvre
enforcement systems designed for a pre-AI criminal landscape. The investment
in international coordination mechanisms — both technical, in terms of
shared AI detection capabilities, and legal, in terms of harmonised
frameworks for algorithmic evidence — is the policy gap whose closure would
most directly improve the odds for law enforcement against AI-enabled
criminal actors. For context on how AI-powered
fraud is affecting individuals at the consumer level, the shadow
economy’s criminal AI adoption has direct household
consequences.

About the Author

Stuart
Kerr is a technology correspondent at LiveAIWire, covering artificial
intelligence, emerging technologies, and their impact on society and
industry.