Google, Microsoft, Amazon and Meta promise to make AI safer and more secure

The White House has received “voluntary commitments” from technology companies that they will help mitigate the risks of artificial intelligence.

US President Joe Biden met with Amazon, Microsoft, Meta, Google, OpenAI, Anthropic and Inflection at the White House on Friday, where they agreed to emphasize “safety, security and trust” when developing AI technologies. Here are some details on each of these categories.

• Security: The companies agreed to “test the security and capabilities of their AI systems, subject them to third-party testing, assess their potential biological, cyber security and social risks, and publicly disclose the results of those assessments.”
• Security: The companies also said they will secure their AI products “against cyber and insider threats” and share best practices and standards to “prevent abuse, minimize risks to society, and protect national security.”
• Trust: One of the biggest compromises was that these companies would make it easier for people to tell whether images are original, modified or created by AI. They will also ensure that AI does not promote discrimination or prejudice, they will protect children from harm and they will use AI to solve challenges such as climate change and cancer.

The arrival of OpenAI’s ChatGPT in late 2022 marks the beginning of major tech companies releasing generative AI tools to the public. OpenAI’s GPT-4 launched in mid-March. It’s the latest version of the Grand Language model that powers the ChatGPT AI chatbot, which is advanced enough to pass the bar exam, among other things. However, chatbots spit out incorrect answers and sometimes resources that don’t even exist. As the adoption of these tools has accelerated, there has been renewed attention to their potential problems, which include spreading misinformation and deepening bias and inequality.

What AI companies are saying and doing

Meta said it welcomes the White House deal. Earlier this week, the company launched the second generation of its flagship language AI model, Llama 2, making it free and open source.

“As we develop new AI models, tech companies need to be transparent about how their systems work and work together across industry, government, academia and civil society,” said Nick Clegg, Meta’s president of global affairs.

In a blog post, Microsoft vice president and president Brad Smith said the deal with the White House “will build a foundation to continue to build on the promise of AI going forward.”

Microsoft is a partner of Meta’s Llama 2. It also launched AI-powered Bing Search earlier this year that uses ChatGPT and is increasingly bringing AI tools to Microsoft 365 and its Edge browser.

The White House agreement is part of OpenAI’s “ongoing collaboration with governments, civil society organizations and others around the world to advance AI governance,” said Anna Makanju, OpenAI’s vice president of global affairs. “Policy makers around the world are considering new laws for highly capable AI systems. Today’s commitments contribute specific and concrete practices to that ongoing discussion.

Amazon supports “As one of the world’s leading developers and implementers of AI tools and services,” Amazon spokesman Tim Doyle told CNET in an emailed statement. “We are committed to fostering innovation on behalf of our customers, while also creating and implementing the necessary safeguards to protect consumers and customers.”

Amazon relies on AI for its podcasts and music, as well as Amazon Web Services.

Anthropic said in an emailed statement that all AI companies “must enter the race for AI security.” The company said it would announce its plans on “cyber security, raid teaming and responsible scaling” in the coming weeks.

“There is a huge amount of security work ahead. Until now, AI security has remained stuck in the realm of ideas and meetings,” Mustafa Suleiman, co-founder and CEO of Inflection AI, wrote in a blog post on Friday. “The amount of concrete progress was insufficient in the face of hype and panic. In inflection, we find this both worrying and disappointing. That’s why safety is at the heart of our mission.”

What else?

Google did not immediately respond to a request for comment, but said earlier this year that it would watermark AI content. The company’s Gemini AI model recognizes text, images and scenes generated by AI. It examines the metadata integrated into the content to let you know what’s unchanged and what’s AI-generated.

Image software company Adobe similarly makes sure to tag AI-generated images from its Firefly AI tool with metadata that indicates they were created by an AI system.

You can read the full voluntary agreement between the companies and the White House here.

The Biden-Harris administration is also developing an executive order seeking bipartisan legislation “to protect Americans” from AI. In addition, the US Office of Management and Budget will issue guidelines for all federal agencies purchasing or using AI systems.

See also: ChatGPT vs Bing vs Google Bard: Which AI Is Most Useful?

Editor’s note: CNET uses an AI engine to help create some of its stories. For more information see this post,