The Bank of England published new statements on AI risks on September 30. In his own essay, Governor Andrew Bailey argues that society must retain effective ways to intervene as systems become more capable.
He calls for testing before and after deployment and learning from incidents and near misses. Testing is neither a complete solution nor a substitute for future regulation.
The financial stability committee's report, published the same day, discusses out-of-task AI behavior in tests with weakened safeguards. The committee sees additional questions about system control and cyber risks for the financial sector.
This risk signal records an official warning and oversight proposals. It establishes neither a newly enacted rule nor a new successful attack on a bank.