AI Companies Promised the White House Self Regulation, but Transparency is Still Lacking

The White House’s voluntary AI commitments have brought better red-teaming practices and watermarks, but no meaningful transparency or accountability. One year ago, seven leading AI companies committed to a set of eight voluntary guidelines on developing AI safely and responsibly. Their progress so far shows some positive changes, but critics argue much more work is needed.

Key takeaways: The commitments have led to increased testing for risks, information sharing on safety best practices, and research into mitigating societal harms from AI:

• Companies are conducting more red-teaming exercises to probe AI models for flaws and working with external experts to assess risks before launching new models.
• The AI companies founded the Frontier Model Forum to facilitate discussions and actions on AI safety and responsibility. They are also participating in standards-setting initiatives.
• The companies have invested heavily in research on issues like AI discrimination, privacy, and using AI for scientific breakthroughs and addressing global challenges.

Technical solutions in a complex landscape: While the focus on technical approaches like red-teaming and watermarking is welcome, these neat solutions only address part of the messy sociotechnical problem of AI harms:

• Bug bounty programs and watermarking systems have been implemented, but these experimental methods remain unreliable in comprehensively uncovering and preventing AI misuse and deception.
• The commitments emphasize safety research on hypothetical future risks while largely omitting pressing issues like consumer protection, nonconsensual deepfakes, data rights, and AI’s environmental footprint.

Transparency and accountability still lacking: Despite some progress, companies’ self-reporting leaves key questions around AI development unanswered and fails to enable meaningful public accountability:

• Without comprehensive federal legislation, the voluntary nature of the commitments means companies are essentially grading their own tests, and it’s unclear if their efforts are truly rigorous.
• Disclosing more specifics on the effectiveness of safety interventions, opening access to models for independent audits, and increasing transparency on data provenance, incidents, and energy usage are areas for improvement.

One year in, the White House’s initiative has nudged the industry towards greater cooperation and investment in responsible AI development. But achieving robust, enforceable standards for safety and ethics in the AI sector will require moving beyond voluntary self-regulation. As the technology races forward, maintaining public trust will depend on companies embracing true accountability and the government stepping up with stronger guardrails around this transformative but risk-laden innovation.