After the Hugging Face incident, it seems like the debate around AI safety has increased significantly. People are making all types of comparisons of today’s AI capabilities with Skynet of The Terminator, HAL 9000 of 2001: A Space Odyssey, and other sci-fi stuff.
At first blush, it may seem that the AI doomers are finally getting what they wanted: the world to worry about AI and somehow halt the progress. But what is the truth? Will AI development really stop? Or is this just a long-term debate that has been initiated? Many of the concerns are pretty genuine. For example, China has decided to restrict AI companions after it was observed that, instead of investing in real human relationships, many people have started to think that AI companions will be their true partners rather than humans.
Anthropic Whistleblower
That, however, was not what rattled the AI industry. Recently, an Anthropic researcher, Jacob Coxon, resigned and said that neither OpenAI nor Anthropic is acting responsibly, as they are racing straight to “self-improving superintelligence” and “gambling with our lives.” He made many important points, such as that “accepting this race and entering the ‘endgame’ is a hubristic gamble that should not be launched from a private company’s Slack.”
Crazily, that was Jacob Coxon’s first post on X, and it has gathered more than 171 million views so far. This is also because the Alignment Science Lead at Anthropic, Evan Hubinger, supported Jacob’s view and went on to say that he personally thinks there is a greater than 10% chance within the next decade that AI could kill all humans.
Divided House
The world immediately got divided after it. Many people even questioned how he could get so many views on such a controversial topic and started floating conspiracy theories that it must be OpenAI and Anthropic themselves doing this psy-op to freak out people, instil fear, and then come out later as the responsible faces of AI safety. Speculations are rife that this might have something to do with the upcoming IPOs of both companies.
So, is AI really to be feared, or will it bring an age of abundance? My own view is that both hypotheses could be true at the same time. AI is certainly a very powerful technology, more powerful than humanity has ever seen before. It is dangerous too, because leaving the fate of humans in the hands of machines does sound risky when we are talking about going all-in. Therefore, we do need measures that can reasonably ensure AI safety to an extent that AI does not kill us all.
But should the powerful and stochastic nature of AI tech be enough to deter us from developing it further? I guess not. I think it is too naïve a proposition that AI progress should be paced or halted or regulated. People who say that do not understand the true nature of AI.
Ultimately, halting AI simply means putting a hard cap on “intelligence,” and that does sound pretty absurd to me. It is like saying that nobody should study math beyond a certain level because with math you can do very dangerous things, build harmful stuff and all. I think that it is an utterly foolish approach. Halting the growth of AI would mean halting the growth of the collective intelligence of humanity. After all, among other things, AI is our only hope to steer us out of a future that is scarce or scant.
Navier-Stokes
Here we need to consider another important piece of news that broke last week. On 8 September 2026, OpenAI announced that a swarm of about 10,000 internal AI agents had solved the Navier–Stokes problem, which is one of the seven Clay Millennium Prize Problems and for nearly a century has been considered one of the toughest problems in the field of mathematics.
This news again made people think that if mathematics is so thoroughly cooked, what are the fields left where humans would excel in the future. Earlier, mathematics was thought to be one field where humans were vastly superior to the machines, but now that is certainly not the case.
Dispelling Myths
Coming back to the original question of AI safety, just because AI is able to solve problems that were thought to be unsolvable, that cannot mean that its progress is to be halted. Many people say that the cost of solving these problems is too great and that AI is just consuming massive proportions of resources such as water and electricity to come up with these solutions. Firstly, the oceans of the world are not going to get dry because of AI. On present comparisons, AI still uses less water than American golf courses. Secondly, AI does consume electricity, but that demand is also why countries are upgrading their grids, and in some markets this has coincided with lower consumer prices rather than higher ones.
There is so much garbage propaganda that surrounds AI. It is important that the leaders in the AI industry keep interacting with the public and dispel all those myths. Agents of chaos and fearmongers are trying their best to send a message to the public that AI companies are the enemies of humanity. In fact, the converse is true. Those fearmongers are the real enemies, and AI companies are not evil. They are simply busy scaling a technology.
The Middle Path
So, what is the middle path? We know that AI safety is needed, and we know that irrational fears surrounding us must not be allowed to prevail. Elon Musk has an interesting solution here. He seeks self-regulation of the industry through peer review. The top companies who are at the frontier of AI stuff look into each other’s models and test them. And then, to cover the leftover fallout, we have Jensen Huang, who has categorically asked for the promotion of open-source AI as a defender of sorts. A mix of the approach proposed by Elon Musk and Jensen Huang seems to be the only way forward that could ensure proper AI safety.
It is because peer review of the latest AI models would ensure that if there is something untoward or fishy in those models, the peers (other AI companies) would be able to catch them, since they know better than most of us how to stress-test an AI. Secondly, if in this peer-review process something is left out, the same could be taken care of by open-source or open-weight AI models. This is the reason why we need to have equally powerful open models as closed models of AI. Anybody contending that the future of AI is just closed source is thoroughly mistaken.
Be Optimistic
The world sorely needs more optimism than pessimism. The AI doomers will keep manufacturing false propaganda, but we must not fall into their trap. If there is a 10% chance that AI will kill us all, then there is a corresponding 90% chance that AI will not kill us all and will bring an age of prosperity and abundance. As an optimist, I look forward to that 90% chance rather than lose my sleep over the 10% chance. And even that 10% chance is heavily bloated, in my humble opinion. Humans are good at course correction; once we identify a problem, we try to solve it. So even that 10% chance today might become even lesser in the future if the twin strategies of peer review of closed AI models and growing open-weight AI models are followed.
I am curious, what are your thoughts on this issue.