The legislation would require AI developers of to report safety failures, cybersecurity breaches, and other incidents to the government.