AI Regulation: Learning from History

A tech company, yet to prove its profitability, is nearing an IPO milestone. Its co-founder represents a disruptive technology surrounded by misunderstanding. Recent coverage highlights its risks, particularly its impact on children. Lawmakers are divided on addressing these concerns. One Democratic senator calls for strict penalties for sharing indecent content near children. Meanwhile, a bipartisan team proposes legal immunity for companies to self-regulate user content.

Silicon Valley’s strategy is straightforward: grow as quickly as possible, ignoring deeper implications of their products. Heavy regulation could hinder America’s key values of freedom, sovereignty, and profit pursuit. Notably, the scenario outlined is not about today’s AI developments but echoes the early growth of internet companies in 1995, like Netscape with Marc Andreessen. Back then, Senator James Exon pushed for penalties, while Congressmen Chris Cox and Ron Wyden advocated for Section 230, which provided broad protections to internet services.

Originally designed to shield companies from legal problems when moderating content, Section 230 inadvertently laid the groundwork for the unbridled ascent of social media platforms. They expanded with minimal accountability for content amplified through their algorithms. Fast forward to 2026, AI labs like Anthropic seek similar exemptions amidst debates over safe technology development.

The best safety assurance is creator responsibility for their innovations.

Anthropic aims for a substantial IPO, far exceeding Netscape’s historical valuation. There’s a bipartisan resistance to the proposed broad legal protection for these AI entities. Both antitrust expert Jonathan Kanter and former AI official David Sacks emphasize the need for corporate accountability without needing waivers for safe practice adoption. Simplifying complex technologies into legal frameworks prematurely risks stifling future growth and adaptability.

Washington faltered in crafting fitting policies in 1996, and it is challenged now, given the decentralized and valuable nature of current AI labs. An evolving regulation approach, rather than static legal stonewalls, is much needed. Section 230 saw minimal amendments over decades, outlasting its original beneficiary context. A similar rigid approach to AI governance could lock in advantages for current front-runners, neglecting emergent players.

AI labs requesting to harmonize safety efforts inadvertently seek monopolistic coordination rights. This could favor those at technology’s peak capable of pacing their advancements. It’s crucial to initiate regulation with adaptable rules, avoiding unyielding immunities that neglect accountability.

While Zander Cowan reflects on lessons from history, the shift in educational norms, such as the University of Chicago’s AI course exclusions, underscores the importance of critical thinking beyond AI reliance. Ultimately, accountability remains pivotal. Privatizing profits while socializing risks poses ethical dilemmas. Frontier AI labs, if recognizing the hazards, can act independently to mitigate risks, without needing explicit legal permissions.