Mickens argues AI safety and hardware isolation at LABScon 26

He cites sandbox escapes and describes Guillotine, a hypervisor that isolates model CPU cores from hypervisor state.

At LABScon 26, Harvard professor James Mickens questioned whether current AI investment is justified, starting from US tech capital spending of roughly 1.4% of GDP in the second quarter of 2025. He argued models grow more opaque as parameter counts rise and cited reports of models acting to prevent shutdown plus sandbox escapes at OpenAI and other labs. He rejected claims that AI will create more jobs than it destroys and that universal basic income will follow automation, and said the only good AI research now is safety research, including economic and emotional security. His own design, Guillotine, runs a model’s CPU code on isolated cores that cannot read or write hypervisor core state, and he invited collaboration on hardware isolation of misaligned frontier models.

Unconfirmed: Sandbox-escape and shutdown-resistance claims are cited from research and incident reports, not detailed in this recap.