Seven Competing Visions of AI Safety Split Tech and Policy Leaders

The phrase "AI safety" has become a political flashpoint, meaning radically different things depending on who's speaking. A new analysis from POLITICO Magazine reveals that the term now functions as a Rorschach test, exposing deep divisions among technologists, policymakers, and advocates over how—or whether—to regulate artificial intelligence.

POLITICO surveyed 20 influential figures across the AI ecosystem, asking each what "AI safety" means to them. The responses clustered into seven distinct factions, each with competing priorities and worldviews.

Why it matters

Understanding these fault lines is critical for business leaders navigating AI policy. Unlike most tech debates, AI safety doesn't break along traditional partisan lines—skeptics and boosters exist across the political spectrum. This fluidity means regulatory frameworks remain unsettled, creating both risk and opportunity for companies developing or deploying AI systems. Knowing which faction holds sway in different policy venues will determine whether your organization faces strict oversight, light-touch regulation, or something in between.

The seven factions

The Doomsday Prophets, concentrated in the Bay Area, believe AI development poses an imminent existential threat. Nate Soares of the Machine Intelligence Research Institute argues the industry is "racing to create artificial superintelligence" with systems creators don't understand. This group advocates for complete development halts.

The Scared Straight Caucus shares existential concerns but stops short of demanding full shutdowns. Philosopher Nick Bostrom focuses on the "technical challenge of aligning superintelligence," while Connor Leahy of ControlAI warns that "without international restrictions on developing superintelligence, an unchecked industry building autonomous AI risks human extinction."

A Safety-Conscious Right faction has emerged among conservatives worried about both catastrophic risks and immediate societal impacts. Mark Beall of the AI Policy Network emphasizes protecting national security from "advanced AI threats such as weaponization and loss of control," while Evan Swarztrauber advocates for child safety protections and export controls.

The Money Movers—philanthropic organizations and advocacy groups with significant funding—take more mainstream positions. Michele Jawando of Omidyar Network frames safety as "a people challenge" focused on systems that "don't concentrate power or spread harm."

Politicians like Senator Bernie Sanders warn that "there are more health and safety regulations around opening up a new sandwich shop than releasing a new AI model," while Rep. Jay Obernolte focuses on preventing "malicious use of AI by humans."

Liberal Believers support AI development with guardrails. Alondra Nelson, former Biden administration science policy official, defines safety around protecting people from algorithmic bias, manipulation, and systems that "erode the institutions and social fabric holding societies together."

The analysis also identifies a Let It Rip Crowd (not fully detailed in the excerpt) representing those who view safety concerns as obstacles to progress.

The effective altruist connection

The "safety-focused" community has traditionally been associated with effective altruism, particularly through its ties to Anthropic. However, even within this movement, significant disagreement exists about optimal regulatory approaches. As the debate has expanded beyond Silicon Valley, the concept has grown increasingly amorphous.

Holly Elmore of PauseAI U.S. exemplifies the internal tensions, calling it "laughable" that Anthropic once positioned itself as an "AI safety" company when it's actually "a frontier AI company."

These details were first reported by POLITICO Magazine in a comprehensive survey of AI safety perspectives across the industry and policy landscape.