SB-1047, known as California's AI Act, mandates that AI developers establish safety protocols, maintain shut-down capabilities for their models, and retain compliance auditors.
The bill aims to prevent generative AI from causing 'critical harm' or exceeding $500 million in losses through cybersecurity failures, setting a regulatory precedent.
Recent modifications to SB-1047 reflect input from AI maker Anthropic, including a shift from mandatory safety test disclosures to less stringent statement submissions.
The act now distinguishes between 'reasonable assurance' and 'reasonable care', altering the obligations for AI developers and thereby impacting liability and compliance expectations.
[
Collection
]
[
|
...
]