Why Biden’s AI Liability Law Faces Resistance From Both Google and Meta at the Worst Possible Time

Even tech executives appear smaller than usual in the Roosevelt Room at the White House because of its subdued weight. Leaders from Google and Meta stood side by side with President Joe Biden in July 2023, nodding courteously as they promised voluntary protections for artificial intelligence. The cameras flashed. They shook hands. For a brief moment, it appeared to be cooperation.

However, the harmony was short-lived. Almost immediately, resistance started to form behind closed doors. The voluntary commitments were symbolic, adaptable, and generally unenforceable. Liability was a completely different matter. Legal responsibility has the power to alter how businesses act and, perhaps more significantly, how they assess risk.

Biden’s proposed liability framework sought to hold platforms responsible for the output of their systems, something Silicon Valley has largely avoided for the past 20 years. This covers algorithmic content, automated judgments, and recommendations produced by AI. It sounds like a technical shift. It isn’t. It targets the core of the contemporary online economy.

It seems that executives are already aware of the stakes as you pass the glass-lined offices of tech companies in Mountain View. Within those structures, engineers hone models that can produce decisions, text, and images more quickly than any human team could. Investors appear to think AI will shape the next phase of corporate power. However, liability causes hesitancy. Additionally, hesitation slows down progress.

Particularly Google has contended that stringent liability laws may stifle innovation, particularly if they are implemented inconsistently across states. The business is concerned about fragmented regulation and a patchwork of laws that force developers to deal with legal ambiguity while foreign competitors are free to operate. That fear might not be wholly unjustified. Policy has always lagged behind technological advancements.

Although they feel a little different, Meta’s worries are just as pressing. There, executives have cautioned that removing liability protections, particularly those associated with Section 230, may leave platforms vulnerable to legal action regarding recommendations made by AI. The business has already fought legal scrutiny over human-generated content for years. Its whole business strategy might change if machine-generated decisions were added to that load.

The irony is difficult to miss.

Both businesses pledged openness and supervision when they signed voluntary safety agreements. However, their tone changed when they were confronted with legally binding consequences. Negotiation replaced cooperation. Resistance evolved from negotiation. Observing this, it seems that voluntary commitments are more easily accepted when they are free of consequences.

Even in Washington, the tension is evident.

Under bright lights and with their voices resonating against marble walls, lawmakers question tech executives inside congressional hearing rooms. Liability is essential, according to some lawmakers. Others are concerned that regulation might make America less competitive. With their technical jargon and political overtones, the arguments go on for hours.

AI continues to develop in the meantime.

Every few months, new models emerge, becoming more competent and convincing. As they race to surpass competitors, companies release updates with cautious optimism, promising safety improvements. The tempo seems unrelenting. Breathless, almost.

For their part, investors don’t seem to be alarmed by regulatory risks. Market values are still increasing. The notion that AI’s economic potential outweighs its legal risks is one that verges on faith. But it’s still unclear if lawmakers and courts will feel the same way.

The ramifications seem more intimate outside of Washington.

People already engage with AI on a daily basis by reading summaries produced by the system, trusting automated judgments, and receiving recommendations. Most people don’t consider liability. If something goes wrong, they believe that someone, somewhere, is to blame.

Soon, that assumption might be put to the test.

Biden has presented AI liability as a public safety issue while he is in the White House. His administration contends that accountability is necessary for powerful technologies. The reasoning seems familiar. Automakers are held accountable for cars that have flaws. Drug manufacturers are held accountable for dangerous drugs. What makes AI unique?

But where Washington sees principle, Silicon Valley sees nuance.

Executives contend that rather than being deterministic, AI outputs are probabilistic. Errors are unavoidable. They contend that liability may stifle creativity and consolidate power in a smaller number of hands. Only giants may be able to afford the risk, as smaller businesses may find it difficult to withstand legal exposure.

There is a curious contradiction in that argument. Businesses oppose the regulation itself while also warning that it could solidify their dominance.

As this debate progresses, it seems more and more apparent that neither side has complete control over the result.

Rarely does technology wait for authorization.

Furthermore, liability rarely goes away once it is introduced.