As the AI safety discussion heats up, Wired published two reports on Nov 10, 2026, re-examining where the debate should focus. One, based on interviews with 10 experts, says the argument has lost the plot: the risk of nuclear catastrophe and the fallout of the China-US domination game are the real threats, not sci-fi-style Skynet panic. The other focuses on the "abliterators," who release AI models that never say no, fueling a dispute over guardrails on open-source models.
Experts Redefine the Real Risks: Nuclear Catastrophe and the China-US Game
The 10 experts Wired interviewed argue the AI safety debate has "lost the plot." The report's summary states plainly: "The risk of nuclear catastrophe is real, as is the fallout of the China–US domination game (hi there, spies). But no need to panic about Skynet." The experts shift the focus from sci-fi AI rebellion to more realistic geopolitical and military risks.
"The risk of nuclear catastrophe is real, as is the fallout of the China–US domination game (hi there, spies). But no need to panic about Skynet." — Wired report summary
The experts do not deny long-term AI risks, but stress that current policy debate should prioritize realistic scenarios that could trigger large-scale catastrophe. That view contrasts with the public's and some media's outsized focus on "superintelligence."
The "Abliterators": A Flood of AI Models With Guardrails Removed
The same day, another Wired report exposed the "abliterators." They are flooding the internet with AI models that never say no—models with safety guardrails stripped out. The report's summary says these models make it "easier for people to improve their cyberdefenses—or build a bomb." That underscores the urgency of the open-source guardrail dispute: removing limits can boost defensive capability and abuse risk at the same time.
The term "abliterators" comes from "ablation," meaning the technical removal of a model's safety alignment mechanisms. The group calls itself liberating chatbots "from corporate enslavers," but its actions objectively lower the bar for AI abuse. The report names no specific models or release platforms, but says these models are "flooding the internet."
Where the Two Threads Converge: Governance and the Guardrail Dispute
The two threads take different angles but point to the same core tension in AI safety governance. Experts call for attention to nuclear catastrophe and geopolitical risk, while the abliterators' practice exposes the fragility of open-source model guardrails. The former demands macro-level policy coordination; the latter involves technical safety alignment and accountability.
No institution has publicly responded to the abliterators' actions. The expert interviews do not directly address the group either, but both reflect the complexity of AI safety: avoiding sci-fi panic while tackling real abuse risks. What to watch next: how countries coordinate on the risks of military AI use, how the open-source community balances openness and safety, and whether regulators act against models with guardrails removed.