Enhancing drug detection in international mail with predictive AI
Project summary
Australia has one of the highest rates of border-controlled drug use globally, with 17.9% of people aged 14 and over reporting use in 2022ꟷ2023. Drug use imposes significant economic costs due to immediate and long-term health impacts. Data from Home Affairs shows that at least 12 tonnes of border-controlled drugs were detected at national mail facilities in the last 10 years.
Australia receives more than 3 million international mail consignments each month, a volume that continues to grow. The detection of prohibited imports at mail gateway facilities has traditionally relied on x-ray examination and manual selection by Australian Border Force officers. While this has led to significant interdictions and the accumulation of valuable intelligence, it is labour-intensive.
Home Affairs has introduced HERMES, an AI-powered model designed to enhance border-controlled drug detection in international mail. The HERMES model marks a shift towards automated targeting and identification of high-risk consignments across the high volume of incoming consignments. This innovation supports border operations with faster, smarter and more scalable threat detection.
International Mail Inspection at the Australia Post Melbourne Gateway Facility.
Image: Melbourne Gateway Facility
How AI contributes to the project
This AI-powered project was developed in-house by the Home Affairs Data Science team to support the detection of prohibited drugs in international mail. It leverages advanced machine-learning techniques and available data assets to enhance targeting for high-risk mail consignments.
The model is built using Gradient Boosting Machine algorithm, a machine-learning method that builds accurate predictions by combining many small learning steps. Each step focuses on improving the mistakes of the previous one, gradually increasing the model’s accuracy. This method is widely used in business for tasks such as identifying risks and improving decision-making.
The HERMES model was trained on a labelled data set to learn patterns associated with drug-related consignments. The Data Science team employed sophisticated feature engineering techniques informed by valuable domain knowledge from Australian Border Force officers. This collaboration ensured that the model incorporated operational insights and real-world targeting experiences.
The model’s design aligns with the principles of the department’s AI Assurance Framework, promoting responsible AI use, transparency, and human oversight. Model outputs are integrated into workflows so ABF officers remain at the centre of decision-making.
Outcomes and next steps
Since its deployment in July 2023, HERMES has delivered impressive results. It has referred around 2,000 international mail consignments, leading to more than 300 detections and stopping more than 400 kilograms of border-controlled drugs from entering Australia. This has helped prevent an estimated $300 million worth of drugs from reaching the community.
The initiative demonstrates the power of AI to improve the efficiency and accuracy of border security operations. The HERMES team received an Australian Border Force Commissioner Award for streamlining threat detection processes and freeing up officers to spend more time on strategic planning, proactive threat hunting and devising mitigation strategies. The team won an award in the Data Science and Analytics category at the APS Data Awards in November 2024.
The development and implementation of this innovative targeting model paves the way for future enhancements in the international mail domain, as well as in cargo and traveller domains. Similar opportunities for integration into existing business practices in these border domains can impact criminal activity and support legitimate transaction facilitation.
The HERMES team is engaging with international partners interested in technical and tradecraft exchanges to support the strengthening of the international mail sector globally.
Find out more
Australian Institute of Health and Welfare (2025) National Drug Strategy Household Survey 2022ꟷ2023, AIHW website, accessed 7 August 2025.
Australian Government (2024) APS Data Awards, APS Data Profession website, accessed 30 July 2025.