As expected and mentioned yesterday, the White House issued a sweeping Executive Order around AI. The EO covers a wide range of concerns. It mandates developers of advanced AIs and large language models to share safety test results, develop standards for AI system safety, prevent dangerous biological materials, combat AI-enabled fraud and deception, establish a cybersecurity program, and draft a national security memorandum.
The order has eight goals: to create new standards for AI safety and security, protect privacy, advance equity and civil rights, stand up for consumers, patients, and students, support workers, promote innovation and competition, advance US leadership in AI technologies, and ensure the responsible and effective government use of the technology.
The order directs leading AI labs to notify the U.S. government of training runs that produce models with potential national security risks, instructs the National Institutes of Standards and Technology to develop frameworks for how to test AI models adversarially, and develop “red team” standards, and establishes an initiative to harness AI to find and fix software vulnerabilities, among other measures automatically. The EO does not require that AI companies adhere to NIST standards or testing methods.
It spells out interim technical standards to assess potential cybersecurity risks of AI models based on computing power and hardware specifications. The order also taps the Defense Production Act to require companies to share AI model testing and training data with government agencies under certain circumstances.
The order requires the Department of Commerce to develop guidance for “content authentication and watermarking” to show clearly labeled marks for AI-generated content. The government will also produce a report on AI’s potential labor-market impacts. It will then outline new principles and best practices to guide employers and workers on job applications and compensation.
The Justice Department and federal civil rights offices will be trained on best practices for investigating and prosecuting AI civil rights violations. Best practices will be developed on the use of AI in sentencing, parole and probation, pretrial release and detention, risk assessments, surveillance, crime forecasting, predictive policing, and forensic analysis.
Administration officials pointed out that the reporting requirement will not impact any AI models currently available on the market, nor will it impact independent or small- to medium-size AI companies moving forward, as the threshold for enforcement is quite high. It’s aimed at the next generation of AI systems that companies like Google, Meta, and OpenAI are already working on, with enforcement on models starting at 10 to the power of 26petaflops, a capacity currently beyond the limits of existing AI models.
President Biden has also requested that Congress pass bipartisan data legislation aimed at validating data collection processes, strengthening research and technology, prioritizing “privacy-preserving techniques,” and developing guidelines for federal agencies to adhere to and evaluate the effectiveness of such techniques.
Finally, the White House launched AI.gov, a new website that showcases the federal government’s efforts and achievements in artificial intelligence (AI) and provides resources and guidance for researchers, developers, and the public.
You can’t say they didn’t do their homework. It’s very clear that AI will not go the way of social media – there will be some level of regulation. That’s very apparent.
The Order delegates to federal agencies to examine AI in their area of expertise. That’s logical and puts the power close to the expertise.
While not directly affecting smaller MSPs or IT outsourcing firms, these standards will likely set industry benchmarks that trickle down, impacting future compliance and service offerings. These are the kinds of frameworks to leverage regardless.
Note that call for privacy legislation. Most Americans believe they already have privacy protection – we covered that data before. It’s possible AI may bring that data protection.
Both of these moves are reasons to help customer make sure their data is well managed and ready for AI… even without the regulation, that’s a requirement to be useful with the technology.