The people who guard the future can't afford to live near it.
The Summary
- Security workers at OpenAI, Anthropic, Google, Meta, and Salesforce authorized a strike, seeking better wages and benefits from staffing agencies Allied Universal, Securitas, and GardaWorld
- SEIU-United Service Workers West represents 14,000 security personnel across San Francisco Bay Area, Silicon Valley, and LA County
- Anthropic already told staff to work from home last week anticipating the strike
- Security threats against AI executives have intensified, including a Molotov cocktail thrown at Sam Altman's home in April
The Signal
The irony is sharp enough to cut. The companies building agents to automate knowledge work can't keep their physical offices open because the humans who secure those offices can't afford rent. Security workers marched through San Francisco Wednesday, past OpenAI and Anthropic headquarters, authorized to strike against the staffing agencies that employ them.
This isn't about AI. It's about the economic structure underneath AI. These workers don't work directly for OpenAI or Anthropic. They work for Allied Universal, Securitas, GardaWorld. The staffing agencies that let tech companies scale security without adding headcount. The same contractor model that's made Big Tech lean also made these jobs precarious.
"The people guarding billion-dollar models live three to a room an hour outside the city."
The timing matters. Security protocols at AI labs have tightened as threats escalated. A Molotov cocktail hit Sam Altman's house in April. He blamed "AI doomerism talk." But the real threat model now includes physical violence, not just prompt injection attacks. Which means these security roles just became more critical and more dangerous.
The 14,000-worker union spans multiple companies and counties. That's rare leverage in a region where tech money usually fragments labor power. If they walk, five of the biggest names in AI and cloud computing go dark simultaneously. Not their servers. Their buildings. The strike authorization doesn't guarantee a walkout, but Anthropic's work-from-home order shows they're taking it seriously.
Here's what the strike reveals about AI companies specifically:
- They've offloaded physical security to the same contractor economy they claim agents will replace
- The wage gap between engineers building AGI and guards protecting them is a chasm
- Remote work is now a contingency plan for labor disputes, not just pandemics
The contract work model shows up everywhere in tech. Content moderators, cafeteria staff, shuttle drivers, now security. Keep the core team small and equity-compensated. Push everything else to vendors. It's efficient until it's not. Until the people at the edges of your operation realize they have collective power at exactly the moment you need them most.
The Implication
If the strike happens, watch how AI companies respond. Do they negotiate directly with workers despite the vendor firewall? Do they replace humans with automated security systems faster? Or do they just wait it out from home, proving that physical offices were always optional?
The broader question: as AI companies automate more knowledge work, what happens to the service economy that keeps their physical operations running? You can't prompt-engineer your way past a picket line. The Fourth Web might be digital-first, but it still runs on people who need to pay rent.