Steven Sinofsky is right: tech labs must not beg for regulation
By Grant Colby · Reporting from Amarillo ·
On the a16z podcast, tech leaders debated federal AI controls, but begging Washington for regulation will only choke free enterprise and leave America vulnerable to foreign adversaries.
On the a16z podcast, tech veteran Steven Sinofsky uttered a warning that every entrepreneur and voter should remember. Once the regulatory wheels start turning in Washington, nobody can slow them down. He sat down with host Erik Torenberg, Box CEO Aaron Levie, and venture capitalist Martin Casado. Sinofsky observed that artificial intelligence has rapidly become an election issue across every level of government. Panicked tech labs are foolishly begging federal agencies to dictate their speed. They worry about public relations and internal fringe factions blathering about species extinction. It is a classic blunder born of corporate naivety. I have balanced payrolls and flown military cargo. I know how Washington operates when handed a new steering wheel. Bureaucrats never tap the brakes to help you. Instead, they build a tollbooth and seize the engine.
The High Cost of Inviting Washington to the Flight Line
As Levie and Casado acknowledged, the strongest case for federal intervention rests on commercial trust and national security. Enterprise customers will not deploy automated software agents if rogue program swarms can hijack internal networks. They will not risk massive data breaches. Furthermore, frontier labs face legitimate anxieties surrounding cyber security vulnerabilities and potential misuse. Proponents of government oversight contend that federal guardrails or pacing agreements provide safety cover. They believe this reassures the market.
That argument falls apart the moment it meets historical reality. Sinofsky rightly pointed out that tech giants from AT&T and IBM to Microsoft made a foolish mistake. They thought playing golf with politicians would shield them from the administrative state. Instead, they got slammed by antitrust lawsuits and endless litigation. Casado brought needed clarity to the hysteria. He noted that if lab executives truly believed AI posed an existential risk of human extinction, they would demand complete federal nationalization. We do that with nuclear weapons at Lawrence Livermore National Lab. They do not actually believe their own apocalyptic rhetoric. Using existential risk to justify regulatory constraints is merely an internal HR gimmick to appease anxious researchers. As Levie noted, regulating technology before understanding its actual mechanics solves nothing while locking in existing risks.
Real Security Demands Engineering, Not Bureaucratic Speed Limits
The historical record confirms that early innovation relies on practical field experience, not premature federal bans. Early aviation operated for decades before formal pilot licensing and airworthiness certifications took shape. This allowed American industry to mature into the safest transport network on earth. We did not strangle the flight line in 1910. Similarly, early computing survived computer viruses and network worms through hard-nosed engineering updates and standardized protocols. It relied on practical authentication layers, not on handing federal bureaucrats a veto over software releases.
Inviting Washington regulators into the AI sector will not make software safer. It will merely insulate incumbent tech giants from Main Street competition while leaving the nation vulnerable. If America stalls its own enterprise through government-imposed speed limits, foreign adversaries will show no such restraint. True security and national strength come from rapid, free-market innovation backed by disciplined engineering. We do not need bureaucratic mandates that choke American enterprise at the gate.