Major Tech Firms Agree To AI Safeguards Set By White House
WASHINGTON (AP) -- Amazon, Google, Meta, Microsoft and other leading AI technology developers have agreed to abide by a set of AI safety measures negotiated by President Joe Biden's administration.
The White House said Friday it has received voluntary pledges from seven U.S. companies to ensure their AI products are safe before launch. Some of the obligations require third-party oversight of how companies operate AI systems, but they don't specify who controls the technology or who holds companies accountable.
A surge in commercial investment in generative AI tools that can convincingly write human-like text and produce new images and other media has sparked public fascination and concern that they can mislead people and spread disinformation, among other risks.
The four tech giants, along with OpenAI, creator of ChatGPT, and startups Anthropic and Inflection, have pledged to conduct security tests "conducted in part by independent experts" to guard against key risks, including biosecurity and cybersecurity, the White House said in a statement.
This test will also explore the potential for social harms such as prejudice and discrimination, as well as the more theoretical risks associated with advanced AI systems that can gain control over physical systems or "reproduce" themselves by making copies of themselves.
The companies have also pledged to develop ways to report vulnerabilities in their systems and use digital watermarks to distinguish between real images and those generated by artificial intelligence, so-called deep fakes.
They will also publicly report on flaws and risks in their technology, including the impact on justice and prejudice, the White House said.
The voluntary pledges are meant to be an immediate way to address risk before a long-term effort to force Congress to pass legislation regulating the technology. Company executives plan to meet Biden at the White House on Friday to pledge to meet the standards.
Some proponents of AI regulation have said Biden's measure is a start, but more needs to be done to hold companies and their products accountable.
“Sincere consultation with business stakeholders leading to voluntary safeguards is not enough,” says Amba Kak, executive director of the AI Now Institute. “We need much broader public consultation and this will raise issues that companies almost certainly won't voluntarily address because it will lead to significantly different outcomes that could affect their business models more directly.”
Senate Majority Leader Chuck Schumer, DN.Y., has announced that he will introduce legislation to regulate AI. He said in a statement that he will work closely with the Biden administration "and our bipartisan colleagues" to move forward on the commitments made on Friday.
Some tech executives have called for regulation, and some went to the White House in May to speak with Biden, Vice President Kamala Harris and other officials.
Microsoft CEO Brad Smith said in a blog post on Friday that his company is making several commitments that go beyond what the White House promised, including support for regulation that would create a "high-capacity model licensing system."
However, some experts and emerging competitors fear that the nature of the proposed regulation could be a boon for well-funded pioneers such as OpenAI, Google and Microsoft, as smaller players are squeezed by the high costs of introducing their AI systems as large language models that meet regulatory restrictions.
The White House compromise notes that it primarily applies only to models that are "generally more capable than the current operating limit" set by currently available models such as OpenAI's GPT-4 and DALL-E 2, as well as similar versions from Anthropic, Google, and Amazon.
Several countries have sought ways to regulate AI, including European Union lawmakers negotiating common AI rules to block 27 nations that could restrict programs deemed most dangerous.
UN Secretary-General Antonio Guterres recently described the United Nations as the "ideal place" to adopt global standards and appointed a body that will report on options for global AI governance later this year.
Guterres also said he welcomed calls from some countries to create a new UN body to support global efforts to manage AI, inspired by models such as the International Atomic Energy Agency or the Intergovernmental Panel on Climate Change.
The White House said Friday it had already discussed voluntary pledges with several countries.
The compromise focuses primarily on safety risks, but does not address other concerns related to the latest AI technology, including the impact on jobs and market competition, the environmental resources needed to build the models, and copyright issues related to human writing, art and other works used to teach AI systems how to produce human-like content.
This month, OpenAI and the Associated Press announced an agreement for the AI company to license the AP news archive. The amount the company will pay for this content was not disclosed.
meter
