The Firebomb at the Door of the AI Age
In the early hours of Friday morning, a young man walked up to the home of OpenAI chief executive Sam Altman in San Francisco's Russian Hill neighbourhood, hurled a Molotov cocktail at the metal gate, and then marched to OpenAI's Mission Bay headquarters to threaten to burn that down too. He was arrested before dawn. Nobody was hurt.
But the attack — crude, frightening, and ultimately unsuccessful — landed like a signal flare over an industry already grappling with questions about what happens when fear of artificial intelligence stops being theoretical.
Daniel Alejandro Moreno-Gama, twenty years old, was booked into San Francisco County jail on Friday afternoon. He faces charges including attempted murder, arson, and possession or manufacture of an incendiary device. Security personnel on site extinguished the fire after the device bounced off the gate, and surveillance cameras captured the full incident, which unfolded at approximately 3:40 a.m.
Police tracked Moreno-Gama down outside OpenAI's headquarters shortly afterwards, where he was making threats. He was taken into custody at the scene.
By the time most of San Francisco had woken up, Altman had already written about it. In an unusual personal blog post published in the small hours of the morning, he described being awake and, in his own word, "pissed." He shared a family photograph — something he said he would not normally do — expressing the hope it might dissuade whoever came next. "I have underestimated the power of words and narratives," he wrote, noting that a recent critical profile of him had been dismissed too quickly.
The post was notable less for its account of the night's events than for what Altman chose to say around them. He acknowledged that fear and anxiety about artificial intelligence is "justified," describing the current moment as "the largest change to society in a long time, and perhaps ever." He called for a "society-wide response" to AI's risks, including new policy to manage what he expects will be a "difficult economic transition." He admitted to being "conflict-averse," which he said had caused real pain, and expressed regret about his handling of the board crisis that briefly cost him his job in late 2023.
It was, by the standards of Silicon Valley CEO communications, unusually candid — and the circumstances that produced it were unusual too.
Investigators and researchers following the incident quickly surfaced a digital trail that suggested Moreno-Gama had been radicalised around AI extinction fears. He appears to have been an active participant in online communities associated with Pause AI, a movement that advocates for halting the development of advanced artificial intelligence on the grounds that it poses an existential threat to humanity. On the group's Discord server, he used the username "Butlerian Jihadist" — a reference to the science fiction series Dune, in which humanity wages war to destroy thinking machines.
In December, he wrote on the Pause AI server that the group was "close to midnight" and that "it's time to actually act." A moderator warned him that calls for violence would result in a ban. The account has since been removed.
Between January and March of this year, Moreno-Gama published six lengthy essays on Substack. In one, titled "A Eulogy for Man," he warned of humanity's extinction at the hands of artificial intelligence and invoked archetypes of "The Warrior" and "The Martyr," writing that some ideas are worth dying for — and killing for. Pause AI has since distanced itself from him entirely.
The group, which has organised protests outside OpenAI's offices and other AI companies, emphasises peaceful demonstration. But the attack has reignited debate about the edge of the AI safety movement and what responsibility, if any, its more extreme rhetoric bears for the actions of individuals who take its warnings literally.
The incident arrives at a moment of heightened tension around OpenAI specifically. The company has been preparing for a high-profile legal battle with Elon Musk, its co-founder turned adversary. It recently completed a restructuring that converted it into a for-profit entity, a move that drew accusations from critics that it had abandoned its founding mission. And Altman himself has been the subject of increasing scrutiny — both inside the industry and beyond it — over his role in shaping the trajectory of artificial intelligence.
For Pause AI and the broader coalition of groups concerned about AI risk, the attack presents a profound strategic problem. The movement's credibility depends on its argument that AI development is dangerous and that those in power are not taking the danger seriously enough. An act of violence carried out in its name — even by someone it has expelled — risks discrediting that argument entirely, handing the industry a narrative in which opposition to AI becomes synonymous with extremism.
None of that appears to have been lost on Altman. "AI has to be democratised," he wrote in his post. "Power cannot be too concentrated. Control of the future belongs to all people and their institutions." Whether that message reaches those who see the institutions themselves as the threat is another question entirely.
Moreno-Gama remains in custody. Altman, by all accounts, has returned to work.
