The rise of AI as a threat to the S&P 500 [pdf]
autonomy.work
Dataset to accompany the report: https://adu.autonomy.work/posts/2025_05_21_ai_risk/
autonomy.work
I'd be curious about the position of these segments in the 10-Ks, like, if they're suddenly all at the top, that's much more interesting than being tacked on to the end.
It's free to put any bad thing in the risks section of your 10-K; investors aren't going to shun your company over it. If you fail to put the risk in, the bad thing happens, and your company loses value, on the other hand, you may get sued for securities fraud - and courts have been oddly receptive to these suits.
It's like any other clause that gets added to any other mostly-boilerplate legal document over time: one firm adds it, pretty soon everyone copies their work and it's a standard term. It's viral. How fast this spreads among company filings is a matter of epidemiology, not something that actually tells you the companies' outlook.
Instagram was moments before ipo.
Also the FB html5 play at the time was a decent bridge to the app play which did of course explode growth.
So let’s be fair they didn’t fail but they acquired to grow even more.
- https://en.wikipedia.org/wiki/The_Interpublic_Group_of_Compa...
- https://en.wikipedia.org/wiki/Omnicom_Group
- https://en.wikipedia.org/wiki/Warner_Bros._Discovery
- https://en.wikipedia.org/wiki/Fox_Corporation
- https://en.wikipedia.org/wiki/Paramount_Global
I'm sure there are a lot more.
What happens when anyone can write a script and have a feature length movie or 12 episode season.
But ooooh boy, would porn change. Maybe that would be the end of pornography, the perfect wish fulfillment making it no longer enjoyably.
Youtube, Facebook, Instagram, TikTok all beg to differ. AI slop farms are making bank over there, especially on short form content aimed at little kids. They get hundreds of millions of views, earning more than what Disney/WB makes on some of their high budget garbage movies in cinema.
What is one specific example?
Oh boy, I hope you're sitting down for this. Here's some people exposing this issue.
https://www.youtube.com/watch?v=2B4_wFOSsuE
I think you're misunderstanding. Consumers don't need to spend money directly, for AI slop farms on Youtube or TikTok to make money, since advertisers pay for the views they get on their videos, not you.
IT's not all about YOU, YOU are probably not the targe audience, but they do make money even if YOU don't watch that stuff.
Where did I say they'll be doomed? I said they're monetizing short form content on YouTube/Tiktok using AI slop, and that can make more money than some crap cinema movies at the box office.
Not when they loose hundreds of millions. Snow White just lost Disney about 150 million.
>A view is worth less than $1.
Yes, and they get hundreds of million of views and failures don't cast anything since they don't need to keep cinemas open.
> This report uses a range of cutting-edge LLM-assisted data techniques to extract key risk information from S&P 500 company filings. Following the recent boom in generative AI, we examine reported risks from these leading firms related to artificial intelligence. We clarify the extent to which firms are reporting new AI related risks, what kind of risks are being reported and what these indicate about the broader dynamics of AI in big business.
This is unrelated to marketing departments.
They want to mention AI to boost interest and such, and they end up mentioning AI risks to hedge/cover their backs
So saying AI is the biggest threat to the S&P is glossing over the root causes. Analysts, getting sucked into the marketing hype, are self-fulfilling in that some people go on their recommendations. After all, in the past if a bank was questioned about how stable it is, could easily see a domino of withdrawals that snowballs into actualy becomming unstable, even if it wasn't before.
I think it doesn't even pass the bar of an undergraduate research paper.
If there is a risk to jobs, it wouldn't show up here since actually less jobs is "good" for business...
Up to a point. Then you no longer have customers.
Prisoner's dilemma, with the businesses as the 'prisoners'.
One of the ways to change the Nash equilibrium for that game is for enough people to empower some outside agent that punishes defectors. (Metaphorically, for the original prisoners in the thought experiment, a gangland boss).
To the extent that the current leadership of government and business are facing a collective action problem it is because different actors have a different number of iterations they are optimizing for.
Put differently, when it's #CrimeSeason, you gotta get yours before the bill is due, and different crooks on different schedules.
But a business needs paying customers, preferably employed by someone else. Other businesses having to pay their employees is good for business in this regard.
The reality is that he couldn’t attract the right workers because factory work is sole destroying.
Hedge funds and investment banks are already using these tools to the max, and the markets are plenty profitable for everyone
I think this scenario is plausible because the path to this scenario is so smooth so it will be the default outcome unless something strange happens to prevent it.
I also don't know that it would change my behaviour. If my goal is long term investment success what's the downside of my continuing to invest in broad market funds? I don't need an AI making split second decisions if my investment horizon is still 30 years.
Consider that millions of people are parking billions of dollars in index funds that just track the ~500 biggest companies with the expectation they’ll all just get bigger.
Everything is Applied Math if you squint hard enough.
All this AI/ML doomerism and boosterism is ridiculous. If you do not understand how gradient descent works or why as of today GPUs are better suited for model training compared to CPUs you should not have a say in this discussion.
Most conversations around AI/ML appear to basically be pop-philosophy discussions that aren't even that grounded in philosophy fundamentals.
If you have trash fundamentals, you will have a trash understanding of the world.
But now think about how most of the population can afford their lifestyle, ie. buying stuff from S&P 500 companies: selling their time as labor.
aside from a few very specific corporate backroom use cases, the pitch for the product currently sold as “ai” is loaded with vague, unclear benefits, marginal utility, and limited real world adoption.
then there’s another product, “ai risks”, many of which are easily identifiable to an average person (layoffs, deepfakes, calculated and miscalculated state violence without accountability), and to the c-suite cause fear and panic, resulting in massive overspending to mitigate these risks, most likely by investing in a “good” ai tool to counter the “bad” ai tool.
how long can the grift continue without any actual positive product to purchase?
It’s a lot more expensive to pay out a securities fraud settlement if you don’t list it and then suffer a loss that can be pinned on ‘AI’.
I look forward to the eventual lawsuit against companies hiding their actual risks in the chaff. Do not use your groin to stop the chainsaw. (To any lawyers, this is the chaff).