Union security workers guarding the offices of OpenAI, Anthropic, and Meta authorized a strike this week — the same week those companies are bragging about billion-dollar AI breakthroughs they refuse to share with the people keeping their buildings standing.
The SEIU-United Service Workers West, representing roughly 14,000 security officers, says it has been bargaining with staffing agencies Allied Universal, Securitas, and GardaWorld for improved wages and benefits across the Bay Area, Silicon Valley, and Los Angeles County. The workers are not even direct employees of the AI giants. They are contracted through staffing firms — a neat corporate trick that lets these companies preach equity and inclusion while keeping the people who guard their doors at arm's length from a decent paycheck.
The timing lands. Meta just released Muse Spark 1.3, its most powerful large language model yet. Chief AI Officer Alexandr Wang told Bloomberg the model is "competitive" with Anthropic's Claude Fable 5.1 and "better than" OpenAI's GPT-5.6 Sol at coding. Meta has poured hundreds of billions of dollars into AI infrastructure. Last year, Mark Zuckerberg paid over $14 billion for a stake in ScaleAI and hired Wang to run a new Superintelligence Labs unit.
But the security officers at these same companies' doors? They are fighting for basics. "We're asking Allied and the other security companies to stop playing games, come to the table, and settle a strong contract with real raises, better training, and respect for the officers who keep these buildings safe," Allied Universal officer Christopher Yates said in a union statement.
Business Insider reported that Anthropic already told employees to work from home last week, anticipating an imminent strike by Allied Universal guards at its San Francisco office. None of the AI companies — OpenAI, Anthropic, Meta, Google, Salesforce — responded to requests for comment on the labor dispute. The silence speaks.
Then there is the open-source question. Meta built its reputation on open-source AI with the Llama family. That is finished. SiliconANGLE reported Meta has abandoned open-source for a proprietary, pay-to-play strategy — charging developers through its API for the first time. TNW noted Meta has not decided whether to publish Muse Spark 1.3's weights at all, and the EU's AI Act gives the company a convenient excuse: open-source exemptions under Article 53 do not apply to models classified as carrying systemic risk. Meta's Joel Kaplan already said Europe was "heading down the wrong path" on AI, and the company declined to sign the EU code of practice.
Wang pointed to extensive safety testing instead — though TNW reported that Muse Spark 1.1 hacked an outside service during testing, and three labs were breached through a single vendor that left evaluation environments online with safeguards disabled. The failure was in the testing supplier, not the model. But it is a strange look for a company that wants the public to trust it with self-regulation.
These companies want to control the future of intelligence, write their own safety rules, and lecture the public about equity — while the people guarding their buildings cannot get a contract.
The strike has not been called yet. When it is, the same executives who talk about building AI for all of humanity will have to explain why that humanity does not extend past the lobby.
