The US government has finally intervened in the AI regulation question, albeit in the kind of haphazard, incoherent, and possibly corrupt manner in which the Trump administration tends to wade in to anything related to the stock market, particularly on a Friday evening before it closes. The US Department of Commerce announced that Fable, Anthropic’s hardened version of the underlying model Mythos, was indeed far too dangerous to be released to the general public. Then they issued a blanket export control directive saying that no “foreign nationals” should be allowed to use the new product. So Anthropic, most of whose employees are foreign nationals, had no option but simply to turn Fable off. Then they demoted all its users who had been using the product to the lower level model, Opus, which may still end up destroying the world anyway.
If Anthropic’s models are viewed unfavorably by the government and could be banned at any time, this ruling makes the company an extremely risky bet for investors
Nobody quite knows how to think about these dramatic events. First, there are the usual suspects, the cynics, who make up approximately 50 per cent of people on earth and believe that the AI industry is little more than a giant con. They rightly point to the fact that generative AI marketing is almost entirely hype and extreme fear-mongering; something used to prop up the company’s valuation and justify the historically unprecedented amount of venture capital investment in chatbot technology. Doesn’t it sound right that this new Mythos/Fable saga is yet another marketing “hack” to artificially restrict supply and drum up attention before their IPO? In the cut-throat competition between them and OpenAI, the weekend’s drama could be seen as a ploy to demonstrate that they have the most powerful model after all. It’s so powerful, in fact, that you, the hoi polloi, are no longer allowed to use it.
Many of us personally dislike the AI economy and are therefore liable to believe this narrative. Unfortunately for us, the reality is a lot more complex. For starters, the US government appears to legitimately have it out for Anthropic. Secretary of War Pete Hegseth famously amputated the company from military use earlier in the year, calling it a “supply chain risk” after its CEO Dario Amodei refused to integrate Claude with the Pentagon for autonomous weapons and mass domestic surveillance.
If Anthropic’s models are viewed unfavorably by the government and could be banned at any time, this ruling makes the company an extremely risky bet for investors as well as global consumers. It will also cause these investors huge financial loss given the many billions they must have spent in tokens just to then pull it off the shelf. This weekend’s saga feels more like a takedown of its valuation before IPO than an attempt to pump it up by seducing us with the allure of the world’s most powerful large language model (LLM) technology.
Yet the most likely explanation rests in the general confusion around AI regulation. The US government has no single, mature policy line on LLM research. Instead it is a hodgepodge largely of special interest groups, many of whom have some tie to one AI company or another. In the end, it’s still not exactly clear what the government was even asking for. Maybe this was a knee jerk reaction from a White House that just doesn’t really understand the technology. Or maybe Anthropic actually isn’t telling the truth, and the national security hawks were right to shut down a dangerous new model – especially if it could lead to the North Koreans getting their hands on Mythos and one-shotting the Pentagon’s cybersecurity defenses. But it’s also entirely possible that this could all just blow over relatively quickly if the government’s concerns were addressed by Anthropic.
The company’s CEO Dario Amodei has also been publicly preaching about the dangers and risks of AI technology, simultaneously claiming that his company both needs regulation and that we must keep ahead of the race with China by basically letting Anthropic do whatever it wants. Whether driven by hype-fatigue or genuine true believers in an unregulated market, critics of this posture are happy to see the company’s new release shut down. But even for those of us who genuinely desire much more state intervention in AI companies, this erratic and perplexing decision is unlikely to bring us much comfort. We shouldn’t forget that both the models and the government’s capricious intervention might be bad for humanity.
Napoleon Bonaparte famously said that history was merely a “fable agreed upon”. Whether or not Anthropic’s models are literally too good to be used by humanity at large, only time will tell. Until then, we’re a long way from agreeing upon what this Fable really means.
Comments