Openness in AI Is Necessary, Said FTC’s Lina Khan

Foundation models, like OpenAI’s GPT4, are capital-intensive, requiring expensive talent, costly computer infrastructure, and volumes of data. “These conditions have allowed, in some instances, the biggest technology companies to get a layup in the AI ​​space,” Khan said.

Incumbents can limit access to critical AI resources, like compute, Khan said. Earlier this year, the FTC launched an investigation into major cloud providers’ multibillion-dollar investments in generative AI startups due to growing concerns about competition. This includes Microsoft’s $13 billion investment in OpenAI, as well as Google and Amazon’s multibillion-dollar investments in Anthropic.

To fully realize AI’s potential, Khan underscores the importance of openness in the field. There needs to be a “real commitment to a philosophy of openness across the industry, which means open markets, but also open architecture, open ecosystems, and open-source software.” Open-source software has historically fostered competition as it allows researchers and engineers to accumulate innovation. In particular, Khan calls for AI models’ weights to be publicly available.

These “open-weight models” lower costs and decreased barriers to entry, enabling innovation at the application layer. However, she warns against the “open to closed playbook,” in which technology may be open-source to start but could reverse course and become proprietary over time.

Khan also notes that the FTC has worked to increase openness in the labor market. “Some of the best engineers in America have been bound by restrictive non-compete clauses,” Khan said. “We’ve heard from startups that secure funding entered the market only to find that they can’t grow because the talent pool has been locked up by dominant players.”

Earlier this year, the FTC banned non-compete clauses, allowing “developers, designers, and researchers across the country to move freely from company to company with their innovative ideas and unique expertise.”

Data privacy is another priority for the FTC, Khan noted. In the past, dominant players have engaged in anti-competitive practices by changing their terms of service or privacy policies and thereby reneging on their user privacy commitments. For instance, the FTC hit Meta with a $5 billion fine in 2019 for misleading users about their data protections.

In the age of AI, Khan said both consumers and enterprises are uncertain of their data’s protections when using foundation models. There is “just a general uncertainty about…whether (firms’) competitively sensitive information and proprietary information is or is not being fed back in.”

“It’s especially important to try to prevent this stuff on the frontend rather than playing catch-up and clean up on the backend when you’ve already lost so much innovation and opportunity,” Khan said.

Antitrust regulators focus only on a “narrow set of deals that satisfy a narrow set of criteria”

The FTC has worked closely with the antitrust division of the Department of Justice.

“I’m a law enforcer, not a regulator,” said Jonathan Kanter, the assistant attorney general for antitrust, at the YC event. “My job isn’t to regulate a market. It is to enforce the law only when companies violate it to make sure competition can thrive.”

In March, the DOJ filed a lawsuit against Apple for maintaining a monopoly over the smartphone market. It also filed a lawsuit against Google, accusing the company of illegally using its dominance to manipulate the search and search advertising markets.

Tech investors have been frustrated by the top regulators’ blockings of mergers and acquisitions. “We can’t sell our companies,” a New York VC recently told Business Insider.

One YC founder expressed in the YC forum that many startups hope to exit via acquisitions, but he said he was concerned about the “chilling effect on acquisitions” due to the DOJ and FTC’s investigations.

However, Kanter highlights that for the vast majority of acquisitions, the DOJ doesn’t do anything. “We focus on the instances where there are bottlenecks to competition,” which is in the single digits, he said. “We only focus on deals that have problems…it’s a narrow set of deals that satisfy a narrow set of criteria.”

To avoid government review, Big Tech has begun exploring new pathways from traditional M&A. In March, Microsoft paid $650 million in a licensing deal to Inflection AI, an emerging OpenAI rival, to use its AI models and hire most of its employees. Inflection’s ex-CEO now heads Microsoft’s AI products and research division. Amazon followed suit, licensing tech and hiring execs away from AI agent startup Adept in June.

Nevertheless, antitrust regulators are still taking notice, launching investigations into both.