The Commonwealth Bank of Australia (CBA) has taken a proactive step in combatting abusive language in digital transactions by open-sourcing its artificial intelligence and machine learning model. The model, implemented on the H2O.ai platform, aids in detecting harassing, threatening, or offensive messages in digital payment transactions, a form of technology-facilitated abuse. CBA has made this AI model available for free on GitHub, offering assistance to banks worldwide in identifying harm in transaction descriptions.
Utilized by CBA to identify approximately 1500 high-risk cases annually, the AI model is an extension of an automatic block filter introduced in 2020. This filter, triggered by threatening or offensive words in transaction descriptions, has already prevented 1 million transactions. In August, CBA announced plans to trial the referral of abusive transaction descriptions to NSW Police with the victim’s consent.
The decision to share the source code with the global community reflects CBA’s commitment to addressing financial abuse and creating a safer banking experience. The initiative, a result of CBA’s partnership with H2O.ai, aims to provide better visibility of technology-facilitated abuse to financial institutions worldwide.
CBA Group Customer Advocate, Angela MacMillan, emphasized the prevalence of financial abuse and its role in perpetuating control in abusive relationships. The technology developed by CBA identifies patterns and instances of high-risk behavior, enabling prompt investigation and action by the bank. By sharing the source code, CBA aims to empower other financial institutions to take informed action against technology-facilitated abuse, ultimately enhancing customer protection.
Other major banks, including Westpac and ANZ, have also implemented technologies to combat financial abuse, reflecting an industry-wide commitment to leveraging innovative solutions to address this critical issue.
