Europol: GenAI Offers “Treasure Trove of Possibilities”

Artificial intelligence (AI) will “profoundly reshape the law enforcement landscape” as long as it is adopted responsibly, Europol’s executive director, Catherine De Bolle has claimed.

De Bolle’s comments came as the policing organization launched a new report into the technology, which claimed that generative AI (GenAI) offers law enforcers “a treasure trove of possibilities.”

The AI and policing report by Europol’s Innovation Lab argued that:

• Generative adversarial networks (GANs) could help police evaluate the performance of biometrics systems without compromising citizens’ privacy, by generating synthetic facial images and other data to assess accuracy
• GANs could also help to combat identity fraud by developing anti-spoofing technologies
• Large language models (LLMs) could empower officers to carry out open source research and intelligence analysis
• LLMs could help teams develop technical investigative tools
• LLMs could speed up administrative tasks like report writing and summarizing information
• Retrieval augmented generation (RAG) could be used to solve the challenge of GenAI hallucinations

“The use of LLMs by law enforcement would require a secure environment that can be trusted with sensitive information, as well as thorough assessments concerning safeguarding fundamental rights and mitigating potential biases,” the report acknowledged.

“Generative AI represents the next leap, moving from passive analysis to active creation. Yet, like any application, its power lies in its judicious and ethical application, balancing innovation with responsibility.”

Among other use cases for AI in policing, the report highlighted:

• Advanced criminal analytics that reveal trends in vast amounts of data
• Natural language processing (NLP) capable of extracting actionable insight from data troves, to improve resource forecasting and operational efficiency
• Real-time insight from open source (OSINT) and social media (SOCMINT) intelligence to tackle urgent crimes like terrorism and crimes against children
• Machine translation to help collaboration across borders

Read more on police use of AI: Europol Warns of Metaverse and AI Terror Threat

Compliance Challenges

The report also acknowledged the constraints of the EU AI Act, which prohibits use of “live” mass facial recognition by law enforcement, with some exceptions. It called for regular audits of AI systems to ensure compliance with privacy and data protection standards.

“The effective development and deployment of AI technologies requires substantial technological infrastructure and expertise, presenting significant challenges, particularly to smaller law enforcement agencies,” it noted.

“Law enforcement agencies must navigate complex legal and ethical landscapes while investing in training and raising awareness amongst their staff to ensure appropriate data handling and responsible data processing practices.”

Additional resources will need to be allocated to help compliance with the EU AI Act, the report continued.

“Accountability, transparency and explainability, are essential not only for ethical and responsible AI use but also to ensure that evidence collected and analysed by AI systems withstands scrutiny, respect the rights to a fair trial and is deemed acceptable in court proceedings,” it said.