On October 23, 2026, Representatives Jay Obernolte and Lori Trahan introduced the 269-page bipartisan Great American Artificial Intelligence Act of 2026, with four additional co-sponsors. It would require frontier developers including OpenAI, Anthropic, and Google to publish safety frameworks, release pre-deployment transparency reports, undergo independent audits twice yearly, and report critical incidents within 15 days—or within 24 hours when risk is imminent. Violations could incur penalties of up to $1 million per day. [1]
The bill defines catastrophic risk as a foreseeable threat causing more than 50 deaths or injuries or $1 billion in property damage, including AI assistance with CBRN weapons, cyberattacks, or evasion of developer control. Its audit regime covers cybersecurity, biosecurity, CBRN uplift, and loss-of-control scenarios; critical incidents include model-control loss and exfiltration of model weights. These categories map directly to autonomous habitat management, spacecraft operations, genomic repositories, industrial systems, and protected civilisational archives. [1]
The Ark team should adapt the bill’s controls into internal AI assurance requirements: maintain public or mission-accessible safety cases, require independent semi-annual audits, classify model-weight theft and control loss as reportable incidents, and establish a 24-hour escalation threshold for imminent threats. Research should also track the proposed federal regulator and its Independent Verification Organizations, while preserving state authority over deployed AI and general laws covering privacy, labour, civil rights, and child safety. [1]