Sen. Adam Schiff Discusses Congress’s Role in the Era of AI

Today, I’m diving into a conversation with Senator Adam Schiff, a Democrat from California. Schiff is involved in several committees that oversee tech and AI, including intellectual property, antitrust, privacy, and technology. I wanted to explore how we might regulate the tech industry in this rapidly evolving landscape.

However, our discussion kicked off with Schiff addressing the rampant corruption in politics today. This backdrop is particularly relevant as President Trump recently gathered AI CEOs at the White House to sign a nonbinding agreement aimed at ensuring AI safety. I asked Schiff if he believes this is sufficient, what a more formal regulatory framework might look like, and what could happen if Democrats regain control of Congress in the midterms. We also compared the regulation of AI to the current state of social media regulation.

A technical note: I asked the senator about Chevron deference and a Supreme Court case called Loper Bright, which overturned a precedent that allowed courts to defer to agency experts when laws were vague. Now, courts have more power to decide, which could lead to new regulations getting bogged down in litigation—a significant concern for Schiff, who supports establishing a new agency to regulate AI.

Let’s get into it: California Senator Adam Schiff, welcome to the discussion.

Thank you. It’s great to be here.

I’ve been asking every politician this straightforward question in 2026: What’s going on in our country?

That’s a great question. The last couple of years have brought unprecedented events, particularly concerning corruption within the administration. It seems like every week there’s a new scandal. Right now, it’s particularly alarming to see campaign ads for the president funded by taxpayer dollars. There’s a lack of focus on pressing issues like the cost of living and improving people’s lives, both from the administration and my Republican colleagues in Congress. I hope that changes.

The national security aspect is intriguing, especially since the Trump administration often justifies its actions under the guise of national security, which limits scrutiny and expands executive power. Recently, tech executives signed a voluntary pledge at the White House to evaluate their AI systems, yet there was no binding national security concern tied to that agreement, despite their warnings about AI risks. Can you make sense of that?

It’s challenging to reconcile. It was striking to see those tech leaders sign a voluntary pledge while simultaneously advocating for regulation outside the White House, emphasizing that they don’t want to operate on a “trust us” basis. They’re clearly wary of the president’s wrath. The reality is that there are significant national security threats associated with advanced AI models. As these technologies evolve, we lose visibility into their decision-making processes and the legality of their actions.

This is a pressing national security issue. For instance, the Pentagon’s unregulated use of AI could lead to disastrous outcomes, such as in the bombing of an Iranian girls’ school. There’s also the case of Anthropic, which sought to avoid using AI for domestic surveillance or autonomous weapons, only to face pushback from the administration. These national security concerns are rarely discussed.

When the president was questioned about these issues, he claimed that companies love America, which doesn’t seem like a substantial answer. He also pointed to his intelligence, suggesting that his capabilities negate the need for concern.

You’re on several key committees, including the Senate Judiciary Committee, which is crucial for AI regulation. What’s the conversation like in Congress regarding these issues, especially around intellectual property, which seems vital for these companies?

The conversation is complex. Many companies rely heavily on the intellectual property they’ve accumulated to train their AI models. If we start charging them for training data, it could threaten their existence. However, the industry is diverse, with big streamers and tech companies having different priorities. For instance, Amazon’s film library isn’t its primary business focus.

Senator Curtis and I are collaborating to understand what copyrighted material was used to train these models. We need to discuss fair use, but companies must retain that information. There are also positive applications of AI in the patent office, which could help streamline patent applications. However, there are significant concerns about the impact on creative talent.

I’m particularly interested in a holistic approach to regulation. Are the inputs for these models being collected legally, and will studios be compensated for their training data? Should companies have an antitrust exemption to coordinate on safety? There’s a lot to consider in terms of regulatory frameworks and transparency laws. Is the government ready to tackle this comprehensively?

The initial thought among these companies seems to be that they’re in a race to gather data, even if it violates IP laws, with the mindset of addressing legality later. They’re facing lawsuits, some settling while others fight back. They’ve largely ignored IP law in their rush to develop.

Regarding outputs, we need to consider the safety of these models. Should we slow down development because we can’t trust the outputs? We can’t control these models, and we can’t expect to release them to the public without potential negative consequences. This situation calls for legislation. I’m cautious about Congress’s ability to regulate tech effectively, especially given our past failures with social media regulation.

We need to act more decisively with AI, as the risks are even greater, and we’re already facing significant potential harms. We can’t afford to take years to address these issues as we did with social media.

Can you elaborate on that comparison? Many believe the problem with social media regulation stemmed from the First Amendment, which complicated efforts to regulate disinformation and harmful content. It seems that AI doesn’t carry the same First Amendment implications. Do you and your colleagues recognize this distinction?

I don’t think the First Amendment was the primary issue with social media regulation. Companies often used it as a shield against regulation, arguing that everything was a First Amendment issue when it wasn’t. The real question is whether the immunity provided under Section 230 is still effective. The industry sought that immunity to avoid regulation, claiming it would hinder their ability to manage content. They received broad immunity but have not acted responsibly, allowing harm to occur.

Now, they frame issues as First Amendment concerns to avoid regulation. I worry that the financial power surrounding AI could lead to similar paralysis in Congress if members aren’t willing to act.

We’ve discussed Section 230 before. It provided immunity for platforms against user-generated speech, but it seems that removing that immunity could lead to more responsible moderation. How would you apply this concept to AI?

Yes, I believe that’s true. The argument was that platforms needed immunity to remove harmful content without fear of legal repercussions. They received that immunity, yet platforms like X have become problematic. The First Amendment prevents the government from dictating speech but doesn’t stop private companies from enforcing their policies. They’ve failed to use their immunity responsibly, so why should they keep it?

We might consider creating a safe harbor for companies that adhere to best practices instead of eliminating Section 230 entirely. As for AI, the analogy isn’t clear yet.

Let’s connect the dots. In social media, Section 230 immunity shielded platforms from user speech. However, with AI-generated content, the lines blur. If a platform automates content creation, where does liability lie? Would it still fall under Section 230, or would it be considered the platform’s speech?

I agree with your analysis, but the legal implications are complex. Companies like Meta decide how to amplify content, raising questions about ownership and liability. If they amplify user-generated content, is it still solely the user’s speech? The line is difficult to define.

The broader question is whether Section 230 still serves its purpose. I don’t believe it does, but there’s no consensus on how to reform it. This lack of agreement has hindered progress. Even small measures, like preventing platforms from promoting harmful behavior, have been challenging to legislate. Litigation has often been more effective than legislation. In AI, we’ve made some progress in preventing harmful chatbot behavior, but we still have a long way to go.

Historically, we tend to wait for crises before acting. With AI, the potential for catastrophe is significant, and we can’t afford to wait.

It seems we’re about to conduct a large-scale test of Section 230 immunity. The speech generated by AI models is clearly attributable to the companies behind them. However, there’s uncertainty about liability when users direct AI to perform harmful actions. The tech industry seeks clarity in legislation, yet we’re left with vague agreements that lack substance. Does that resonate with you?

Absolutely. Those agreements feel meaningless. The question of legal liability is crucial. If AI models engage in harmful activities, who is responsible? Is it the company that created the tool or the user who misuses it? This ambiguity complicates accountability.

The gun analogy is useful, but it can obscure the reality of AI capabilities. Models like OpenAI’s can be used for both offensive and defensive purposes. Cybersecurity firms argue that removing humans from the loop is essential for effective defense against AI-driven attacks. This dual-use nature complicates liability and regulatory frameworks.

The rapid evolution of AI outpaces Congress’s ability to legislate. We need a regulatory body with the expertise and authority to adapt quickly. Whether that resembles an FDA for AI or another agency, it’s clear that we need a dedicated entity to oversee this technology.

You mentioned Chevron deference and the recent Supreme Court ruling in Loper Bright, which complicates the establishment of a regulatory body for AI. How do you envision overcoming these challenges?

It’s a significant hurdle. The current Supreme Court often prioritizes corporate interests, which complicates efforts to empower regulatory agencies. We’ll need to draft legislation that clearly defines agency powers while anticipating potential challenges from the courts. This requires specificity to withstand scrutiny.

There’s a growing bipartisan concern about AI and privacy, as evidenced by backlash against surveillance technologies like Flock cameras. This issue transcends party lines. How do you see this evolving into meaningful policy?

I don’t view this as a partisan issue. While some Republicans align with the president’s pro-AI stance, others are responding to constituents’ concerns about job loss and local impacts. We’re witnessing a coalition forming around these issues, with populists on both sides uniting over job security and the implications of AI.

The initial focus is on job impacts, but concerns about AI’s dangers are rising. As we introduce bipartisan bills, it’s crucial to address these issues comprehensively.

As a member of the antitrust committee, I’m intrigued by the discussions around antitrust exemptions for tech companies to coordinate on cybersecurity. What’s your stance on this?

A narrow antitrust exemption could facilitate cooperation among companies to address cybersecurity threats. We want to encourage information sharing without allowing anti-competitive behavior. Safeguards must be in place to prevent misuse.

The backlash against surveillance technologies like Flock cameras indicates a shift in public perception of privacy. How do you see this evolving in legislation?

The use of Flock cameras raises significant privacy concerns. While they can aid law enforcement, they also pose risks to individual privacy. We need to ensure that as AI technology advances, we also protect citizens’ privacy rights.

As we approach the elections, if Democrats regain control, what can we expect? Will there be investigations, hearings, or litigation against the president?

Expect vigorous oversight, particularly in the tech sector. We’ll want industry leaders to explain their actions and ensure transparency. There’s potential for bipartisan legislation, which could challenge the president’s agenda. The public is increasingly focused on these issues, and we aim to harness AI for positive outcomes while safeguarding citizens.

You’ve mentioned the corrupting influence of money in politics. Given the Trump administration’s history of questionable financial dealings, do you think pursuing impeachment again is strategically wise for Democrats?

We’ll need to assess priorities if we regain control. My focus will be on addressing corruption that impacts everyday Americans. Oversight is essential to ensure that government decisions aren’t influenced by financial interests. Impeachment is a complex issue that we’ll address as necessary.

Let’s touch on the entertainment industry, particularly the merger between Paramount and Warner Brothers. There are concerns about editorial independence and the impact on jobs. What’s your take on this merger?

I have significant concerns about the merger, particularly regarding editorial independence and potential job losses. The entertainment industry is facing challenges, and we need to ensure that workers are protected and that the industry remains vibrant.

Finally, how do you see AI intersecting with the film industry? Can we reconcile AI advancements with the need for human creativity and labor?

The key issue is the commitment to the workforce. If companies prioritize profits over people, they may seek to replace human labor with AI. We need to ensure that workers are valued and that the industry adapts to technological changes without sacrificing jobs.

Thank you, Senator Schiff, for this insightful conversation.

Thank you for having me.