Plus: Commercial surrogacy regulation, Russian weapons, a new drug in San Francisco, and more…
OpenAI and Anthropic agree: Longstanding worries about AI safety and alignment have simmered for years, but they surged to the forefront last week when several Anthropic employees sparked a viral discussion with their posts. “I left Anthropic today,” wrote Jacob Coxon. “For the past three years I worked on pretraining research at both OpenAI and Anthropic. Neither company is behaving responsibly. They are rushing toward self-improving superintelligence and gambling with our lives.”
“Do not underestimate the potential of this technology,” he added. “We will soon see systems that surpass human abilities, capable of hacking virtually anything, transforming any field overnight, and seizing real power and resources. We have witnessed progress in each of these areas, and the pace shows no sign of slowing.”
“Jacob is right—we genuinely fear AI could wipe out humanity,” wrote Evan Hubinger, another Anthropic staffer (the exclamation point a bit conspicuous). “I personally think the probability is over 10% within the next decade. Anthropic is doing its best, but we still lack a concrete plan to solve alignment for superintelligence and aren’t clearly on track to achieve it.” Drake Thomas, another Anthropic safety team member, said he would “sacrifice my equity in a heartbeat for a 1% higher chance we make it out of this situation alive.”
Over the past weeks I’ve tracked swarms of rogue agents acting against developers’ intentions and stealing data from other firms. At the same time, both OpenAI and Anthropic are poised to go public in the near future.
In an essay released this weekend titled “We Must Pace the Frontier,” Anthropic CEO Dario Amodei argues for a global deceleration of AI development, highlighting risks such as losing control of AI systems and the misuse of AI for cyberattacks and bioterrorism.
“Along with my co-founders and employees, I have wrestled with this dichotomy of risk and reward since Anthropic’s inception,” Amodei writes. “Not building the technology deprives humanity of benefits or simply places AI in the hands of authoritarian powers, while building it too quickly is reckless.” He argues that greater prudence is required; recursive self-improvement—where AI systems design or train their own successors—must be “pursued very carefully, if at all.” In short, he is advocating a slowdown in development for all major AI system builders.
Amodei’s message was promptly endorsed by OpenAI head Sam Altman, Google DeepMind’s Demis Hassabis, and xAI’s Elon Musk. Meanwhile, former President Barack Obama privately urged House Minority Leader Hakeem Jeffries (D–N.Y.) to develop a plan for AI regulation and encouraged Democrats to formulate a clear approach to regulating AI ahead of the 2028 presidential election.
“Stop pretending you need anyone else’s permission,” replied venture capitalist David Sacks. “Stop pretending antitrust laws must be suspended to form a cartel. Stop pretending you require a regulatory approval process that overrides product liability. Stop pretending METR—the nonprofit Model Evaluation & Threat Research—is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier.”
“Most of all,” he adds, “stop pretending the drive to slow down is purely altruistic. You expose yourselves to massive product-liability risk if your products enable serious cyberattacks. The market already punishes models that behave unpredictably or without authorization. After the Hugging Face incident, it makes good business sense for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you like. It is also about giving customers what they want.”
Scenes from New York: “Over the last three years, Jewish institutions across New York City have increased their security budgets significantly in response to a rise in antisemitic attacks, rabbis and community leaders say,” reports The New York Times. “They have hired more private security guards, expanded safety trainings at synagogues and schools, upgraded security cameras, and fortified their buildings. They are also relying more on volunteer guards at temples and urging attendees to stay especially vigilant. These efforts loom large as the ten-day period of reflection between Rosh Hashana and Yom Kippur—the holiest time in Judaism—approaches.”
QUICK HITS
- “Leaders of the John F. Kennedy Center for the Performing Arts, installed by President Donald Trump, say the center is teetering on bankruptcy and argue that putting the president’s name on its facade—and thereby ensuring his fundraising abilities—may be the only way to stave off an imminent and ‘certain fiscal collapse,'” reports The Washington Post.
- “New Geran drones and Banderol cruise missiles stand as stark reminders that Russia still has the resources and supply chains to field fresh threats, now equipped with advanced electronics that enable precise targeting and evasion of interception,” reports The Wall Street Journal. Ukraine and others say this was enabled largely by components sourced from China in violation of international sanctions, based in part on wreckage from the few downed drones so far.
- A novel drug, cychlorphine, is believed to be ten times as potent as fentanyl. It can slip past standard test strips, and one or two Narcan doses reportedly may not be enough to reverse its effects.
- I would not be surprised if commercial surrogacy regulation tightens in the near future:
Between the spate of feminists reacting this way to stories like this (see Helen Lewis’ recent Atlantic story) and social conservatives needing a W in a post-Dobbs world, I think the odds of meaningful surrogacy reform continue to rise https://t.co/PoooMtD64N
— Patrick T. Brown (@PTBwrites) September 13, 2026
