Washington: Leading artificial intelligence (AI) tech companies in the United States including Google, OpenAI and Meta have reached a deal with the Joe Biden administration to roll out fresh guardrails to manage risks associated with AI.
The top seven companies are Amazon, Anthropic, Meta, Google, Inflection and OpenAI. The companies formally announced their commitment to new standards in the areas of safety, security and trust during a meeting with President Joe Biden at the White House on Friday afternoon.
The measures would include testing the security of AI and making the results of those tests public.
"These commitments are real, and they are concrete. AI is going to change the lives of people around the world. The people here will be critical for shepherding that innovation with responsibility and safety by design," Biden said at the White House following the meeting.
"AI should benefit the whole of society. For that to happen, these powerful new technologies need to be built and deployed responsibly," said Nick Clegg, Meta's president of global affairs.
"As we develop new AI models, tech companies should be transparent about how their systems work and collaborate closely across industry, government, academia and civil society," he added.
As part of the agreement, the tech companies have agreed to security testing of their AI systems by internal and external experts before their release.
This will ensure that people are able to spot AI by implementing watermarks and publicly reporting AI capabilities and limitations on a regular basis.
These companies will also research the risks such as bias, discrimination and the invasion of privacy.
"This is a serious responsibility, we have to get it right. There's enormous, enormous potential upside as well," said Biden.
OpenAI said that the watermarking agreements would require the companies to "develop tools or APIs to determine if a particular piece of content was created with their system."
Google has committed to deploy similar disclosures earlier this year.
Earlier this week, Meta said that it would open-source its large language model Llama 2, making it free for researchers, similar to OpenAI’s GPT-4.