Michaël Trazzi

I organized the biggest AI Safety protest in US History! Nearly 200 people marched from Anthropic to OpenAI to xAI with one demand: commit to pausing if the others do too

@jackclarkSF At what point will Anthropic pause automation of AI R&D / recursive self-improvement R&D, so that it can focus on making sure RSI doesn't lead to catastropic outcomes, including human extinction? Would Anthropic pause frontier AI development if China & other labs also did?

I've spent the past few weeks reading 100s of public data sources about AI development. I now believe that recursive self-improvement has a 60% chance of happening by the end of 2028. In other words, AI systems might soon be capable of building themselves.

The judge from Musk v. Altman instructed Musk to not talk about extinction again

Google is apparently keeping their best coding model secret to focus on accelerating AI R&D internally Only Anthropic can access Claude Mythos to automate AI research This is not how humanity wakes up to the dangers of recursive self-improvement. This suicide race has to stop.

Following our March 21st Stop The AI Race protest, Sam Altman agrees that pausing AI development may be necessary in the future "We expect there will be periods where we need to collaborate with governments, international agencies, and other AGI efforts to ensure that we have sufficiently solved serious alignment, safety, or societal problems before proceeding further with our work."

@imjmcinnis There you go washingtonpost.com/technology/202…

In The Washington Post this morning

In-depth coverage of AI Safety Activism on NBC News

Met with @tedlieu's office in DC. He breaks AI risk into three buckets: AI we're not worried about, AI that could hurt people, and AI that could kill everyone. He signed the 2023 CAIS statement and I've been impressed by how seriously his office takes AI's extinction threat.

Stop The AI Race mentioned in The Atlantic

Regarding recent events

Regarding the molotov cocktail, I told the New York Times: "We condemn all violence. The Stop The AI Race movement is and always will be lawful and nonviolent." Also note that we want the pause of *frontier* development. Full Article: nytimes.com/2026/04/10/us/…

Our demand of asking conditional pause statements from major AI Company CEOs is in the New York Times today

@ShakeelHashim Just use protest pictures! x.com/MichaelTrazzi/…

i cannot express enough how hard it is to find art for AI stories

In the New York Times this morning

@ben_j_todd I do get the perspective of "let's try to maximize our chance of things going well by making sure enough resources go to Alignment research in the next two years" But if the plan is to get to a treaty, taking action before the labs get exponentially more power is urgent

@ben_j_todd AI companies are gaining an increasing amount of power every day, automating their own AI R&D, increasing their revenue exponentially, etc. The question should be "How are you planning to prevent bad outcomes in the coming months", not "how will you spend the next two years"

You may have two years left to prepare for the intelligence explosion. How are you going to spend them?

@robbensinger Why are there three Sam Altman?

Who should I add to this? Also, did I get anyone's view wrong?

Overall, I want to mention that I think Kai (@chi_t_williams) was a pleasure to talk to, asking questions that really challenge the people he interviews. This is also one of the only pieces doing an in-depth analysis of AI Pause activism across different movements. So I recommend reading his piece & following more of his work! understandingai.org/p/bernie-sande…

Most importantly, Kai covers the protest in detail but doesn't mention that Anthropic dropped its commitment to pause development (in RSPv3) weeks before we marched on their headquarters, which was in our press release and on our website, and the reason we started at Anthropic.

Kai also mentions a couple speakers focusing on chatbots encouraging teen suicide instead of focusing on existential risk. But if we can't stop a chatbot from killing a teenager, why would we trust these companies with superintelligence?

There also seems to be convergence on the instrumental goals: moratoriums on data centers (Bernie, datacenter groups & AI Safety people), OpenAI having less power (AI Safety, QuitGPT), stopping the race (coalition at this protest). You don't need to agree on why to agree on what.