Sam Altman’s San Francisco home has endured two attacks within days, part of three incidents connected to AI tensions. The violence started with a Molotov cocktail on April 10, 2026, and ended with gunfire over April 12-13. These events expose raw public frustration with AI progress.
The Sequence of Attacks
On April 10 at 4:12 a.m., 20-year-old Daniel Alejandro Moreno-Gama hurled an incendiary device at the gate of Altman’s North Beach or Russian Hill property, valued between $27 million and $65 million. The fire stayed outside, and no one got hurt. Moreno-Gama ran off but matched a police description. By 5 a.m., he showed up at OpenAI’s Mission Bay headquarters, threatening to set it ablaze. Officers arrested him there. Prosecutors hit him with eight felonies, from arson on an occupied building to attempted murder.
Records show Moreno-Gama posted 34 messages over two years in the PauseAI Discord server, a group against advanced AI like future GPT versions. He wrote pieces titled “Eulogy for Humanity,” ranting about AI extinction risks. PauseAI banned him after the arrest and rejected any connection to his actions. The group pushes for halting frontier models but disavows violence.
Two days passed, and on April 12-13, a car pulled up to the same residence. Shots rang out toward the house. This marked the second strike on Altman’s property and the third AI-related assault in that brief window. The pattern runs from the Molotov at Altman’s door, the HQ threat, to this shooting. Suspects Amanda Tom, 25, and Muhammad Tariq Hussein, 23, got caught after police traced a license plate. Officers found firearms in their vehicle. No injuries occurred, but the rapid succession raises alarms about escalating threats to AI executives.
Responses from OpenAI and Altman
OpenAI put out a statement after both events: “Early this morning, someone threw a Molotov cocktail at Sam Altman’s home and also made threats at our San Francisco headquarters. Thankfully, no one was hurt. We deeply appreciate how quickly SFPD responded.” The company kept it short, focusing on gratitude to law enforcement.
Altman went further in his own blog post. He tied the attacks to a recent New Yorker piece that stirred up AI worries. “I underestimated the power of words and narratives,” he wrote, adding that he felt pissed off now. Altman called for toning down heated anti‑tech talk, pointing to how stories fuel real‑world anger.
Media handling drew criticism too. Outlets like the San Francisco Standard printed Altman’s full street address, even after the first attack. Altman owns several spots in that North Beach and Russian Hill area, but listing specifics invites danger. This practice questions how newsrooms balance public interest with personal safety for high‑profile targets.
Roots in AI Discontent
These strikes stem from wider unease about AI’s rise. PauseAI and similar outfits demand a stop to giant models, fearing job losses, privacy breaches, and worse. Moreno‑Gama’s Discord activity shows how online spaces amplify those views, even if groups pull away from violent acts.
Surveys back the shift. One poll found 55 percent of Americans see AI as harmful, up 11 percent from last year. That number climbed as models like GPT‑5 variants rolled out, handling everything from code to complex reasoning. Over 140 groups now rally against AI, targeting data centers and offices at OpenAI and Google. Protests hit streets, with chants about extinction risks and economic fallout.
I’ve covered how fast AI moves forward, from the leaked OpenAI Spud details to benchmarks pushing coding scores past 77 percent. That speed creates a gap between what labs build and what people grasp. Fears aren’t baseless—AI already shifts jobs in writing and design, as I’ve noted before. But violence crosses a line that debate should handle.
Implications for AI’s Path Forward
What comes next tests the field. Opposition to AI won’t vanish; the question is its form. Right now, it tips toward direct action against leaders. Altman and teams at OpenAI, Anthropic, and Google must boost security while tackling root concerns. Simple statements won’t cut it—they need clear plans on ethics, job impacts, and transparency.
Take model releases: Claude Mythos hit high marks on software engineering tasks, and Spud promises more. These gains drive value, but they also spark backlash when seen as unchecked power grabs. As I argued in pieces on math capabilities debates and code tools like Claude Code, capabilities grow, but so does scrutiny. Public views on AI mirror that—polls show distrust in ads and ethics, with 45 percent of younger folks wary.
Media plays a role too. Dropping addresses isn’t just sloppy; it endangers. Outlets should redact such details for figures under threat. Broader, AI firms could fund better discourse—forums, not just press releases—to counter narratives like those “Eulogy” essays that push doomsday scenarios.
This wave of attacks signals AI’s societal footprint deepens. Labs focus on benchmarks, but real progress means handling human reactions. Ignore the backlash, and incidents multiply. Address it head‑on, with facts over fear, and the field stabilizes. The core issue stays: AI builds tools that reshape work and thought. Response demands balance, not dismissal.
Looking at my takes on agents replacing routine tasks or open‑source lags, the pattern holds. AI delivers, but at a cost to perceptions. These events force a reckoning—how do we advance without alienating half the population? That answer shapes the next releases, from Spud to whatever follows.