Microsoft, Google, OpenAI, and other companies to sign an agreement with the White House on AI development principles
According to Bloomberg, today representatives of the leading US artificial intelligence companies will publicly commit to the White House on the security of their technologies.
Microsoft, Google, OpenAI, and others are expected to commit themselves to the responsible development and implementation of AI. Although the commitments are voluntary, they are the result of efforts by the Biden administration and Congress. The U.S. Parliament spent months holding hearings on the issue, which has caused significant disagreement among congressmen. Companies’ obligations will end when Congress passes a law regulating AI.
AI technology developers have previously warned about the risks of its use. The Biden administration, for its part, has offered guidelines for AI development, including the National Institute of Standards and Technology’s “Risk Management Framework,” which emerged from months of interaction with industry leaders and experts.
In a document likely to be released today, the White House will propose eight commitments focused on security and social responsibility. They include:
- Allowing attempts by independent experts to push models to behave badly (a process known as “red teaming”)
- Sharing trust and security information with government agencies and other companies
- Using watermarks on audio and visual content to identify content created by AI
- Investing in cybersecurity tools
- Encouraging third parties to disclose security vulnerabilities
- Communicating the social risks of AI, such as misuse and bias
- Prioritizing AI social risk research
- Using advanced artificial intelligence systems known as boundary models to solve society’s most serious problems
So far, the efforts of governments and companies have lagged far behind the pace of AI development, driven by fierce competition between corporations and fears that Chinese AI technologies could overtake Western achievements.
