Asian regulators heighten cybersecurity scrutiny over Anthropic's Mythos AI model
Regulators across Asia stepped up scrutiny of cybersecurity risks in financial systems following concerns about Anthropic's Mythos AI model. Anthropic reportedly held back wider release after finding the model could discover previously undetected security holes, prompting regulatory responses in Singapore, South Korea and Australia.
What happened
Anthropic held back a wider release of its Mythos AI model after finding it was capable of discovering security holes that have gone undetected for years.
Singapore's financial regulator urged banks to plug holes, while South Korea's government agencies met to review and discuss how to respond to the risks. In Australia, authorities expected lenders to be vigilant to ensure clients were not put at risk by inadequate controls.
Reported impact
The capabilities of the model fueled alarm about a potential new era of cybersecurity attacks. Lee Won-tae, chairman of South Korea's special committee on security, stated: 'We have entered an era where AI, not humans, holds the initiative in hacking.'
Index analysis
The model's ability to discover previously undetected security holes across years suggests potential for exploitation at scale against critical infrastructure. A South Korea special committee suggested setting up an AI-based real-time defence system as part of the response.
Governance controls
Anthropic limited release of the Mythos model due to identified cybersecurity risks. Regulators across Singapore, South Korea and Australia initiated scrutiny and coordination with financial institutions to assess and mitigate risks posed by the model.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.