In the fast-paced world of frontier technology, a young company gearing up for a significant initial public offering faces challenges similar to those of the past. A co-founder takes center stage, representing a technology misunderstood by many. Recent reports highlight the risks associated, especially concerning children. The government remains divided on solutions. A Democratic senator suggests jail time for uploading indecent content near children, while a bipartisan group of congressmen advocates for immunity from lawsuits over user posts, allowing companies to self-regulate.
Silicon Valley operates with a straightforward strategy: prioritize user growth and avoid scrutinizing product or user activities. Strict regulations posed by the government threaten America’s core values, including freedom, sovereignty, and the pursuit of profit. This scenario mirrors the internet’s emergence in 1995, with Netscape and Marc Andreessen. The Democratic senator was James Exon, and the congressmen were Chris Cox and Ron Wyden, co-authors of Section 230 of the Communications Act.
This legislation stated, “No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.” It aimed to address a business dilemma: Dial-up services faced lawsuits for moderating content. The creators didn’t anticipate that this shield, meant for self-regulation, would become the foundation of social media, allowing platforms to expand unchecked, even when harmful content emerged.
By 2026, Silicon Valley seeks fresh privileges: an antitrust exemption and protection from lawsuits related to system and user-generated content. Recently, Dario Amodei penned a widely shared essay advocating for these. On September 15, Treasury Secretary Scott Bessent told a House committee that these labs shouldn’t receive liability exemptions. “The best way to ensure safety is for creators to be accountable for their outputs,” he asserted.
Meanwhile, Anthropic has plans for an October IPO valued at $2 trillion, about 700 times Netscape’s worth. Thankfully, bipartisan efforts are challenging these demands. Jonathan Kanter, Joe Biden’s antitrust head, informed CNBC that frontier labs could embrace safety standards without antitrust relief. David Sacks, a former AI leader under President Trump, emphasized firms’ responsibility to ensure product safety.
Sacks’ viewpoint aligns with historical lessons. In 1996, Washington couldn’t legislate the best policies, and it struggles to do so now. The last deep-seated technological shift, the Manhattan Project, saw the government controlling atomic labs, knowledge, and access decisions. In contrast, today’s private AI labs eclipse most economies. The market would not logically hand them over to the state.
The House adjourned for the midterms, leaving outdated AI bills pending. Expecting Congress to lead is akin to asking an untrained passenger to steer a bus to safety. This complex technology requires adaptive regulations, not fixed pledges. There’s no opposition to AI regulation here, only caution against establishing unchangeable rules during an industry’s infancy.
Section 230 has been amended once in three decades. This shield initially designed to protect a few dial-up services outlasted its creators. A similar exemption or waiver today might crystallize around current industry giants. AI labs seeking self-coordination on safety seek authority to cooperate, and only frontier firms can afford to decelerate.
Regulation should start, but with flexible guidelines that evolve alongside the technology. Avoid granting unchangeable immunity. Unlike young Andreessen, Amodei, now 43, can learn from history. Recently, the University of Chicago barred AI from foundational social science courses, urging students to embrace traditional literacy and critical thinking.
These antiquarian texts impart a vital lesson: Shields endure beyond their protectors, binding survivors to what they shielded. Accountability is integral to profit pursuit. Privatized gains shouldn’t coexist with socialized risks, nor with unchecked permissions for potential pitfalls. If AI labs perceive their innovations as perilous, they can proceed more cautiously without external consent.