AI Companies Make ‘Voluntary’ Safety Commitments at the White House

Concerns about the potential risks and lack of transparency in the rapidly expanding field of artificial intelligence (AI) have prompted the Biden administration to take action. To address these issues, the White House has convened seven of the most prominent AI developers to discuss common goals of safety and openness. The voluntary pledges made by these businesses are nonetheless seen as a positive development towards the establishment of regulations and guidelines for the AI industry.

Representatives from the following artificial intelligence companies were present at the White House meeting:

1. Brad Smith, President of Microsoft;
2. Kent Walker, President, Google;
3. Dario Amodei, CEO, Anthropic;
4. Mustafa Suleyman, CEO, Inflection AI;
5. Nick Clegg, President, Meta;
6. Greg Brockman, President, OpenAI;
7. Adam Selipsky, CEO, Amazon Web Services;

Although the AI companies’ pledges are not legally binding, they do show a desire to address the issues that have been raised about AI. The major promises made by the businesses are as follows:

1. Before releasing their AI systems to the public, the companies have committed to conducting internal and external security tests. Experts from outside the company can act as an adversarial “red team” to probe for security flaws and threats.
2. Companies using AI have promised to inform policymakers, academics, and members of civil society about the dangers posed by AI and the methods being developed to counteract them. Concerns like preventing unauthorized access or “jailbreaking” of AI systems are among those they hope to address by encouraging collaboration and transparency.
3. The companies will invest in cybersecurity measures, including protections against insider threats, to ensure the privacy of client model data and intellectual property. This is essential to thwart hackers’ attempts to exploit security flaws and steal private data.
4. Through bug bounty programs and/or domain expert analysis, the companies will encourage third-party discovery and reporting of vulnerabilities. This promotes outside inspection, which is useful for discovering vulnerabilities in AI systems.
5. Companies are working on developing reliable watermarking or other methods of marking AI-generated content to ensure accountability and traceability. This is crucial in the fight against disinformation and deepfakes, as it will allow for the identification of the original creator and authenticity of AI-generated content.
6. Companies have pledged to disclose their AI systems’ strengths and weaknesses as well as the contexts in which they should and should not be used. This openness is crucial for preventing misuse of AI and ensuring its responsible deployment.
7. The companies will put a premium on studying the social risks of AI, such as discrimination and invasion of privacy. By identifying and mitigating these threats, they hope to create AI systems that are equitable, apolitical, and protective of individual privacy.
8. The companies will also work on creating and deploying AI to tackle some of society’s biggest problems, like stopping cancer and fixing the climate. The lack of monitoring of AI models’ carbon footprint, however, calls attention to the need for more eco-friendly considerations in AI research and development.

The AI companies’ pledges are entirely voluntary, but the Biden administration is working on an AI-related Executive Order. In its current form, this order has the potential to promote compliance and set industry standards. The Executive Order could instruct government agencies like the Federal Trade Commission (FTC) to investigate AI products claiming robust security, for instance, if companies do not permit external security testing of their AI models prior to release.

The administration’s forward-thinking stance on AI reflects its commitment to avoiding technological pitfalls in the future. Because of the lessons it has learned from social media’s disruptive potential, the government is eager to launch a comprehensive plan for artificial intelligence. Vice President Harris and President Biden have already consulted with business moguls and other influential figures. There has also been a substantial investment in AI research institutions and initiatives.

The national science and research infrastructure, however, has been ahead of the government in this respect. There is already a comprehensive report outlining the difficulties and potential benefits of AI for science that was compiled by the Department of Energy (DOE) and National Labs.

Finally, the major AI companies’ voluntary commitments made at the White House are a big deal for ensuring safety and transparency in the AI industry. Although the pledges are not legally binding, they do show a commitment to working together, sharing data, and putting an emphasis on ethical AI research and development. There needs to be a middle ground between being innovative and making sure AI is used in a responsible and ethical way.

First reported on TechCrunch

Frequently Asked Question

Q. What was the purpose of the White House meeting with AI developers?

The meeting aimed to address concerns about potential risks and lack of transparency in the AI industry. The Biden administration convened seven prominent AI developers to discuss common goals of safety and openness in AI development.

Q. Which companies were represented at the White House meeting?

The AI companies represented at the meeting were Microsoft, Google, Anthropic, Inflection AI, Meta, OpenAI, and Amazon Web Services.

Q. Was there diversity in the representation at the event?

No, there was a lack of diversity in the representation at the event, as no women were present among the AI company representatives.

Q. Are the AI companies’ pledges legally binding?

No, the pledges made by the AI companies are voluntary and not legally binding. However, they demonstrate a commitment to address issues raised about AI and promote responsible AI development.

Q. What are some of the major promises made by the AI companies?

The companies have pledged to conduct security tests before releasing AI systems, inform policymakers about AI dangers, invest in cybersecurity measures, encourage third-party discovery of vulnerabilities, disclose strengths and weaknesses of AI systems, and work on addressing social risks of AI, among other commitments.

Q. How will the Biden administration promote compliance and set industry standards?

The Biden administration is working on an AI-related Executive Order that could potentially instruct government agencies, such as the Federal Trade Commission, to investigate AI products claiming robust security if companies do not permit external security testing of their AI models before release.

Q. How does the administration’s stance on AI reflect its commitment to avoiding pitfalls?

The administration’s forward-thinking stance on AI shows its dedication to learning from past technological disruptions and proactively launching a comprehensive plan for AI. It involves consulting with business leaders and investing in AI research institutions and initiatives.

Q. Is there existing research on AI’s potential benefits and challenges?

Yes, there is already a comprehensive report compiled by the Department of Energy and National Labs outlining the difficulties and potential benefits of AI for science.

Q. Are the AI companies committed to ethical AI research and development?

Yes, the companies’ voluntary commitments at the White House show their dedication to working together, sharing data, and emphasizing ethical AI research and development to ensure safety and transparency in the AI industry.

Originally published on ReadWrite.