Directives from the order include:
- Mandatory sharing of safety test results for AI models that could impact national security, economic stability, or health.
- The introduction of guidelines for “red-team testing”, where testers mimic potential malicious actors.
- Official protocols for watermarking AI-generated content to counteract risks from fraud and deepfakes.
- Development of standards for screening biological synthesis to prevent AI-assisted creation of bioweapons.
Jeff Zients, White House chief of staff, emphasized the urgency, quoting President Joe Biden: “We can’t move at a normal government pace. We have to move as fast, if not faster than the technology itself.”
I’m skeptical that government is going to move as fast as AI, let alone faster than AI, when it comes to regulating the technology.
I believe the White House is sincere in the need to regulate AI, but this order comes with other members of the G7 and EU also releasing their AI regulations. Feels a little politicized, which I guess everything is.
Further provisions encompass:
- Labeling and watermarking AI-created deepfakes to differentiate genuine interactions from software-generated ones.
- Addressing privacy, civil rights, consumer protection, and workers’ rights.
- Directives for the military and intelligence sectors on ethical AI use.
- Urging Congress to establish laws safeguarding American data privacy.
- Offering guidance to prevent AI-induced discrimination in housing, federal benefits, and contracts.
- Establishing best practices for using AI in the judicial system and addressing job market disruptions.
- Encouraging the Federal Trade Commission to monitor potential AI market distortions.
It’ll be interesting to see how tech and government agencies work together on AI. Feels like tech is not only ahead in understanding how AI works and will be developed, but also has a vested interest in AI from a profits side.