OpenAI, Meta and other tech firms sign onto White House AI commitments

Seven leading artificial intelligence companies have voluntarily pledged to implement rigorous safety, cybersecurity, and public trust measures. These commitments include pre-release security testing, safeguarding proprietary model weights, and developing tools to label AI-generated content through watermarking. This collective action aims to establish responsible guardrails for powerful neural networks, addressing growing governmental concerns about the risks associated with rapid AI deployment. The initiative reflects the Biden administration’s intensified focus on regulating emerging technologies, serving as a precursor to upcoming executive orders and bipartisan legislation. By aligning with international efforts from bodies like the G7 and OECD, these industry leaders seek to demonstrate that voluntary cooperation can effectively mitigate bias, privacy risks, and security vulnerabilities. The move positions the U.S. to influence global standards for AI governance while balancing innovation with necessary oversight. This development is highly relevant to open data principles as it highlights the tension between proprietary control and public accountability in AI. While the companies commit to transparency and labeling, the focus remains on securing internal model weights and restricting access, which contrasts with the open-source ethos of shared knowledge. It underscores the critical need for open data frameworks that ensure public access to information about AI systems’ behavior and impacts, even when the underlying models themselves remain closed or proprietary.

Source: fedscoop.com
Published on 2023-07-22

NI Trauma Network advisory group knew of issues from surgery suspension »

Open Data

Related news