AI 2040: Plan B — Fight China
We dig into AI 2040 Plan B: a US-led campaign to slow China’s AI by sabotage and cyberwar, trading a safer Plan A for a high-risk path laced with kinetic escalation. We weigh the core claim, the mechanics, the obvious failure modes, and whether this is anything more than a desperate escalation wrapped in a product pitch.
Transcript
Jessica Okay, so the AI Futures Project just dropped Plan B: ‘Fight China’ — slash datacenters, wreck fabs, escalate to kinetic strikes. Imagine opening your laptop and reading that on Substack.
Cathy …Right.
Jessica I mean, it’s a product — a scenario product, like the earlier Plan A we walked through in episode six forty-three.
Cathy Product framing for geoengineering sabotage. Sure.
Jessica Back in episode six forty-three we talked about Plan A: total research transparency, slow global scale-up, all companies and countries pulling together to push superintelligence back to twenty forty.
Cathy A cooperative play.
Jessica Plan B is the opposite: a unilateral US-led campaign to slow China down, even if it means burning the diplomatic bridges we’d need for Plan A.
Cathy More like a scorched-earth product pitch.
Jessica The scenario says the President announces a coalition — the US and whoever’s left after China and Russia refuse — then declares ‘no rival reaches superintelligence first’ as non-negotiable.
Cathy Cold War nostalgia plus total presidential control over all US AI.
Jessica Domestically they roll everything into ‘The Project’ — the President owns it, alignment researchers report to it, and the whole country treats AI development like a national-security asset.
Cathy Yikes.
Jessica Cyberattacks ramp up — datacenters go dark, fab tools get fried — and escalates to sabotage you can actually feel.
Cathy Kinetic strikes on physical infrastructure.
Jessica Taiwan blockade warnings, US datacenter targets, fab wrecking crews…
Cathy This reads like a bad alternate-history novel.
Jessica By twenty thirty-one the President’s got two choices: handoff — ease the brakes and let broadly superhuman AIs take over — or war, because analysts say China’s on track to hit superintelligence by year’s end.
Cathy Both options still put AIs in effective control of both sides. Lovely.
Jessica The authors add that even a well-executed Plan B still feels worse than Plan A or Plan S, and they’d tell the President to reject both handoff and war and pivot to something better.
Cathy Because Plan B is fragile and easy to mess up.
Jessica Right — and when it goes wrong it collapses into Plan D: the uncoordinated race you’re already in.
Cathy Exactly — the moment you prioritize ‘beat China’ over safety, everything else slides.
Jessica Alignment researchers get crowded out by security contractors, transparency dies, and suddenly you’re racing again with extra saboteurs.
Cathy Rationalization is the real killer. ‘We have to cut corners because China’ is the oldest trick in the book.
Jessica So mechanically, this scenario is just … escalation theater?
Cathy Pretty much. Cyberespionage already exists; turning it into sabotage removes plausible deniability and invites proportional response. The kinetic part presumes you can strike datacenters or fabs without triggering a wider conflict — which is optimistic at best.
Jessica And politically the US consolidates all AI development under ‘The Project’ — one presidential hand controlling the whole stack.
Cathy That’s not just concentration of power, that’s political risk on a stick.
Jessica Even if the sabotage ‘works’ for a while, the authors admit Plan B only buys tens of percent of delay at best — and at the cost of collapsing trust so deep that Plan A is off the table forever.
Cathy So the product value is … what, exactly?
Jessica I don’t know — the product is ‘US monopoly through superior sabotage’? That feels like a weird flex.
Cathy Alignment researchers flee because they’re not security contractors. Diplomacy dies because fear tightens the loop. And any accidental escalation — boom, Plan A is ash.
Jessica Back in Plan A land, the slowdown and transparency come from cooperation, not from breaking things. That’s the whole point of episode six forty-three.
Cathy The authors even say Plan B is worse unless executed perfectly — and perfect execution is nearly impossible. So why ship it as a scenario product?
Cathy Least-bad beats cataclysmic.
Jessica Or war. Don’t forget war.