--- title: "How I Came to Take AI Risk Seriously" description: "My path from dismissing AI risk to working on AI safety, and the effort to think clearly amid fear, social pressure, and uncertainty." published: 2026-09-13 tags: ["Learning and research practice", "AGI strategy and foresight"] importance: 4 cover: "https://jacquesthibodeau.com/content/images/2026/09/how-i-came-to-take-ai-risk-seriously-header.jpeg" coverAlt: "An open notebook on a lakeside porch, with spruce silhouettes and a dark storm gathering across still water, rendered as a textured print." audio: "https://pub-4ee2f71bc29541a7a6e8d9694f0a1b21.r2.dev/how-i-came-to-take-ai-risk-seriously/audio.mp3" author: "Jacques Thibodeau" canonical: "https://jacquesthibodeau.com/how-i-came-to-take-ai-risk-seriously/" --- As an AI safety researcher, here’s my journey into believing risks from advanced AI were serious: In 2015, someone suggested I should pivot my career into addressing the risks. I looked into it and didn’t think it would be a big deal. Then, I spent time looking into forecasts of malicious use of AI in 2018: [The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation](https://arxiv.org/abs/1802.07228) At the same time, I looked at AI/ML use in countries like China to understand how it could be used to influence the population. Over time, it became clear that AI would be moving much faster than I (and many others) expected. But even then, I heard what AI safety people were saying throughout 2020-2021, and it didn't feel real. I didn't really buy it and sort of worried these folks had fallen into some intellectually status-seeking groupthink. Mind you, this was way before the sky-high valuation and public status of being an AI researcher at a top lab. The thing is, I never *really* engaged with the arguments or studied the technical details as to why it would be hard to align a superintelligent AI. I was still largely reactive and perhaps unconsciously didn’t even want to engage properly because it could materially change all my life plans for the worse (not to mention seeming like a crazy guy worried about some AI future). So, in January 2022, I decided to give this a real shot and immerse myself in the technical challenge. Luckily, I had done tons of introspection during the pandemic and could enter it clear-eyed. After studying from rural Canada, I went to Berkeley for two months in the summer of 2022 to do research. I was reticent to go because, even though I saw the alignment challenge as a serious issue at that point, I didn't want to get absorbed in the social scene in a way that would keep me from engaging with a sober mind. As a result, despite being surrounded by smart, cool people, I didn’t engage much socially. I was still at a stage where I could be roped into the social dynamics of believing in AI risks; I didn’t want that to influence me. For similar reasons, I decided I shouldn't work at a frontier AI lab. It would be too easy to rationalize continuing to build even if, deep down, I was terrified of how wrong things could go. Years have passed since then. Many of us never thought the world could take AI this seriously. We thought it wouldn’t wake up until it was too late. Crazy AI scenarios were just in the movies. Partially as a result, many other folks did take those jobs at AI companies. I had many discussions with some of those OpenAI and Anthropic employees on Twitter in Berkeley back in 2022. Some of them were more worried than I was; some still seemed unsure. Immersed in the problem day and night, I had my first moment of terror that summer in Berkeley. And then, I made peace with it. I had to. I needed to be strong and have courage for my loved ones. Since then, I've made a conscious effort to engage with those who believed the risks were overblown. In most cases, it was clear that these people had simply not engaged with the technical research in any serious way. Their arguments were largely rationalizations for some glorious AI future they envisioned. Now, don’t get me wrong, I had plenty of technical and value-based disagreements with those who engaged with the AI risk research too. Lots of this stuff is confusing and complex! Ultimately, several of them went on to work at the labs because they wanted to push for safety from the inside. I think many of them are surprised about how much the world has woken up to the risks and how many early warning shots we’ve been able to get. I’m sure some felt they were screaming into a void for years. Others resigned themselves to working on the technical challenge from the start because they felt it was hopeless to do anything else. And now here we are. Many are waking up to the risks, but many others are dismissing them. It’s a bit messier than we thought, but most of us, on either side, just want what’s best for the future of humanity. ## Sources Every external link in this piece that has a captured card, with what that page said when it was captured. The quoted lines below are not the author of this piece writing: they are the linked page describing itself, recorded by `bun run link-cards` on the date given, and kept so that a reader still has them if the original moves or goes away. - **The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation** — arxiv.org Captured 2026-09-13. > This report surveys the landscape of potential security threats from malicious uses of AI, and proposes ways to better forecast, prevent, and mitigate these threats. After analyzing the ways in which AI may influence the threat landscape in the digital, physical, and political domains, we make four high-level…