The White House has accepted voluntary agreements from leading AI startups, including OpenAI, Alphabet, and Meta Platforms, to implement safety safeguards such as watermarking AI-generated content.
Microsoft, together with Anthropic, Inflection, Amazon, and OpenAI, made a commitment to rigorously test systems before making them available as well as to share knowledge about ways to lower risks and invest in cybersecurity.
The decision is considered a victory for the Biden administration’s efforts to control the technology, which has witnessed a boom in investment and consumer appeal.
Since generative AI—which uses data to create fresh material like ChatGPT’s human-sounding prose—became so popular this year, legislators everywhere started thinking about how to lessen the risks the technology poses to national security and the economy.
The seven companies agreed to work together to create a system to “watermark” all types of content, including text, photographs, audio files, and videos created with AI, so that consumers will be able to tell when the technology has been used.
The technical embedding of this watermark in the content should make it easier for users to recognise deep-fake visuals and audio.
Additionally, the firms committed to concentrating on user privacy protection while AI is developed, as well as ensuring that the technology is neutral and not used to discriminate against weaker groups.
