Why are the regulators saying “Use frontier AI, but have a plan”?
APRA and ASIC are urging financial markets to be prepared, implement strong governance, and build resilience, urging them to shift from merely raising awareness of frontier AI risks to taking decisive action.
Let’s understand the objective behind using frontier AI and how the financial landscape is gaining from it.
What is Frontier AI, and how is it different?
Frontier AI refers to the most advanced AI models, with capabilities far beyond those of current AI. These models can reason, code, generate images, perform scientific tasks, and solve complex problems.
Frontier AI generally has the following capabilities.
- General capabilities– They don’t perform a single function and are built for multiple tasks
- Strong reasoning capabilities– They can help with research, follow multi-step functions, and assist with programming and analysis.
- Cross-format ability– They can understand text, images, and videos and also create them
- Tool use– They can interact with external tools, browse information, execute code, or use APIs when enabled.
As APRA has highlighted, the use of AI in decision-critical, customer-facing, and embedded operations is no longer experimental; let’s see a few of the many finance-related functions where Frontier AI will play, or already plays, a major role.
- In financial analysis, they are used to analyse financial statements, identify future trends, perform comparative analysis and generate results, which are used by management to make strategic decisions.
- In financial reporting, they are used to prepare reports by extracting information and reconciling it.
- For credit assessment, which includes assessing loan applications and other related activities.
- To support investment decisions by comparing market information, company reports, and economic data.
- The regulatory requirements are compared against documents and transactions to ensure compliance is followed in the process.
In recent months, APRA and ASIC have been organising roundtables, with participation from the Australian Signals Directorate, the Reserve Bank of Australia, the Treasury, and the Australian Competition and Consumer Commission.
These discussions by APRA and ASIC emphasise the necessity of following various guidelines.
- The importance of getting the cybersecurity fundamentals right
Establishing strong identity and access controls for critical assets and systems is necessary. If damage happens due to cybersecurity issues, the response and recovery arrangements need to be in place. Backup integrity and 3rd-party risk management should be verified.
- Consider key decisions at board level
The timeframe available for firms to detect and respond to cyber threats has notably narrowed due to the speed at which frontier AI operates. So, key decisions on risk appetite, escalation authority, recovery priorities, and communication strategies should be made and implemented even before the crisis hits.
- Growth of defensive AI
As the cybersecurity risks related to AI increase, the technology to defend against them should also grow. Threat intelligence, early vulnerability detection, and incident response are also on the rise in this scenario.
- Common dependency and concentration risk
Any risks associated with third-party service providers can lead to major operational disruption.
- Contributing to industry-led collaboration
The threats posed by frontier AI can be addressed more easily if businesses contribute more to industry-led collaboration. This includes sector-wide threat intelligence sharing, dependency mapping, supplier assurance, and sector incident coordination.
The biggest concern for financial regulators isn’t frontier AI itself, but its extraordinary capacity to act, scale, and impact more quickly, which can turn a minor vulnerability into major financial disruption.
When used with secure techniques, proper governance, resilience measures, and protocols in place, and when they are well prepared and implemented before any cyber strike occurs, frontier AI enhances processes across financial services.
Source: APRA
