Amazon, Google, Meta, Microsoft and other tech firms agree to AI safeguards set by the White House

President Joe Biden revealed on Friday that several tech firms including Amazon, Google, Meta, Microsoft, and others are backing new artificial intelligence (AI) safety guidelines developed by his administration. These commitments, Biden stated, are vital for managing the considerable potential and risks associated with AI.

Biden’s administration has gained voluntary pledges from seven U.S. firms, including OpenAI, Anthropic, and Inflection, designed to ensure that their AI technologies are safe prior to launch. Some of these commitments include the involvement of third parties in examining the functioning of next-gen AI systems, though the specifics regarding auditing and accountability remain undefined.

The President asserted that there is a “fundamental obligation” for these firms to ensure the safety of their products, considering the potential threats associated with emerging technologies. He cited the damaging impact of powerful technology lacking adequate safeguards, as evidenced by social media, and claimed that while the commitments represent a positive step, there’s more work to be done.

Public interest and concern have grown with the increase in commercial investment in generative AI tools. These tools, capable of producing human-like text and generating new media, carry risks such as spreading disinformation and deceiving people.

According to a White House statement, the companies pledged to conduct security testing, partially handled by independent experts, to guard against significant risks like biosecurity and cybersecurity. Other potential harms to society, such as bias and discrimination, and theoretical dangers like advanced AI systems gaining control of physical systems or self-replicating, will also be examined.

The firms also promised to establish methods for reporting vulnerabilities in their systems and using digital watermarking to differentiate between real and AI-generated content, also known as deepfakes.

In a meeting behind closed doors on Friday, executives from these firms assured the President of their commitment to uphold these standards. Inflection CEO Mustafa Suleyman emphasized the significance of this collaborative endeavor, considering the competitive nature of the industry.

Under the agreement, the companies will also publicly disclose flaws and risks associated with their technology, particularly regarding fairness and bias.

The aim of these voluntary commitments is to address imminent risks while advocating for the development of legislative regulation over a more extended period. However, some proponents of AI regulations believe more needs to be done for corporate accountability.

Senate Majority Leader Chuck Schumer (D-NY) intends to introduce legislation to regulate AI, in conjunction with the Biden administration’s efforts and the commitments made by the firms.

Microsoft President Brad Smith indicated his company’s support for regulation that would establish a licensing regime for highly capable models.

Concerns are rising among experts and budding competitors that the proposed regulations could be advantageous for large-scale first movers like OpenAI, Google, and Microsoft, leaving smaller players at a disadvantage due to the cost of regulatory compliance.

While the White House pledge addresses safety risks, it does not touch upon other issues concerning AI, such as the impact on jobs and market competition, the resources required to build the models, and copyright issues related to AI learning from human-created content.

OpenAI recently struck a deal with The Big Big News to license AP’s archive of news stories. The financial details of the agreement were not disclosed.

This report comes from O’Brien, based in Providence, Rhode Island.

Frequently Asked Questions (FAQs) about AI Safety Guidelines

More about AI Safety Guidelines

• Artificial Intelligence: Safety, Security, and Society
• White House Technology Policy
• Understanding AI Technology
• The Future of AI Regulation
• Voluntary Commitments for AI Safety
• AI and Cybersecurity
• Artificial Intelligence and Bias