What AI Tells Us About Benefit Corporations

Look, AI may kill us all but it will generate some fantastic headlines along the way. To wit:

Anthropic researcher believes more than 10% chance AI ‘could kill all humans’

A.I. Could Possibly End Humanity. How Are Humans Supposed to Process That?

AI staff complain of mental toll over fears of threat to society

How should investors position for the robot apocalypse?

and the chef’s kiss:

Tech leaders to UN: For the sake of humanity, please control the AI technology we created

The part that’s funny here, of course, is that the tech leaders who are raising the alarm that AI could destroy humanity are completely in charge of making sure that doesn’t happen. I seem to recall Sam Altman being fired – and leading a revolution for his reinstatement – over exactly that fear.

Now, in an ordinary business corporation, one might say – legitimately – that the managers have a fiduciary duty to maximize wealth for their shareholders, and therefore cannot let pesky concerns like the obliteration of humanity factor into their considerations, at least not if, taking those concerns into account, they still end up with a positive net present value.

Of course, that fiduciary obligation cannot be enforced in any real way, which is to say, no shareholder could sue an AI board for overindexing on safety while failing to maximize future profits, but boards could be forgiven for taking that obligation seriously nonetheless, and therefore putting out a call to the world’s leaders to alter the legal rules – somehow – to countermand their corporate law instructions.

But OpenAI and Anthropic are not ordinary corporations. They are both benefit corporations, specifically so that their boards can be relieved of the legal obligation to maximize profits, and are permitted to make the judgment call that the destruction of humanity is not worth the increase in shareholder value. The entire justification for taking control of these entities away from shareholders – and housing it, in both cases, in a nonprofit entity – is that these guardrails are necessary to ensure responsible AI development.

So it’s rather ironic to hear tech leaders insist that these protections are, essentially, fruitless when pitted against the profit motive.

Now, to be fair (as I previously posted) one possible argument is that the benefit corporation form is inadequate to constrain the profit motive for industries (like AI) that require extensive capital investment. But it still begs the question why OpenAI and Anthropic need to take all that control away from shareholders (which they surely will continue to do once they are publicly traded).

In any event, the real issue here seems to be something like this:

And another thing. New Shareholder Primacy podcast! This week, we have another epic crossover event with our sister pods Business Pants and Proxy Countdown, to discuss the SEC’s proposal to rescind Rule 14a-8. One additional point on this: We mention on the show that there are already comments up at the SEC website. Several of those are asking for more time beyond the 60 day comment period. This has become a theme; the SEC proposes huge amendments, and commenters on all sides ask for more time given the radical nature of the changes, and the fact that the SEC has not apparently examined how they interact. In this case, the interaction effect is subtle but important. For example, take the proposal to allow semi-annual reporting. If shareholders have access to 14a-8, they can communicate in a systematic way with management about their preferences for reporting cadence; without 14a-8, that becomes much more difficult.

Anyhoo, here at Apple; here at Spotify; and here at Youtube.