Sam Altman touched down in Washington this week. The OpenAI chief executive scheduled meetings with Trump administration officials and lawmakers. His agenda mixes briefing on the company’s latest model with urgent talks on AI oversight. Yet the visit marks something larger. It highlights how two fierce AI rivals have found common cause inside the Beltway.
Altman’s trip comes as the administration races toward an Aug. 1 deadline. That date caps a 60-day window set by a Trump executive order to shape a voluntary review process for advanced models. OpenAI has already given feedback on the draft framework. So have Anthropic and Google. The process aims to let companies submit models for government-backed safety checks before release. OpenAI still believes in the need for some kind of national standard passed through legislation, said Chris Lehane, the company’s chief global affairs officer, in Semafor.
But the coordination runs deeper. OpenAI and Anthropic have quietly teamed up on policy matters even as they battle for customers and talent. Their joint push centers on mandatory evaluations for powerful new systems. Both warn that unchecked open-weight models could hand dangerous capabilities to adversaries or bad actors. The alignment surfaced in recent weeks as the firms lobbied the Trump administration for industry-wide reviews. One recent X post captured the moment bluntly: “OpenAI and Anthropic lobby Trump administration for industry-wide model reviews.”
Spending tells its own story. Anthropic nearly tripled its federal lobbying outlays in the first half of 2026. The company reported $3.53 million through June, adding the Treasury Department to its target list. OpenAI spent $2.22 million in the same period. In the second quarter alone the pair poured a combined $3.17 million into federal advocacy, a 23 percent jump from the first quarter and a fresh high for both. They targeted issues from cybersecurity and copyright to cloud computing and defense procurement. Figures come from federal disclosures analyzed in CNBC and echoed across Quartz.
The numbers dwarf earlier years. They also outpace many legacy tech players who dialed back their own Washington efforts. And the outlays reflect more than defense. Both labs see regulation as a strategic moat. Tighter rules on frontier systems could slow smaller competitors and open-source projects while giving established players time to iterate safely. Critics call it self-serving. Supporters call it responsible. The debate rages on social platforms and in closed-door sessions alike.
Anthropic has taken a particularly aggressive stance on safety. CEO Dario Amodei has repeatedly flagged catastrophic risks from open-weight models. He insists on nuanced rules for the most powerful systems rather than outright bans. Yet the company refused to join a recent industry letter signed by Nvidia, Microsoft, Meta, Google, OpenAI, Palantir and others urging the administration not to restrict open-weight AI. That decision lit up X. One post framed it as a deeper fight: whether intelligence becomes something humanity owns or something it must rent forever.
At the state level the two companies diverge in tone but converge in ambition. OpenAI’s top lobbyist Chris Lehane coined “reverse federalism” to describe its strategy of pushing similar AI bills across states to create a de facto national standard. The approach bypasses a gridlocked Congress. Anthropic, by contrast, presses for ever-tougher safety requirements in state capitals. It endorsed a Massachusetts frontier safety bill that demands independent testing every six months for catastrophic risks, public reports, and attorney general enforcement powers. OpenAI backed a milder version of the same legislation. Details emerged in Politico and Axios.
Both firms also face questions about their own data practices. Court filings revealed Anthropic’s “Project Panama,” an industrial-scale effort launched in early 2024 to buy millions of print books, cut their spines with hydraulic machines, scan the pages at high speed and discard the physical copies. An internal memo described it as “our effort to disruptively scan all the world’s books” and warned “we do not want it to be known that we are pursuing this project.” The company later settled a related copyright lawsuit for $1.5 billion. The account appeared today in news.com.au. Such episodes feed skepticism about the labs’ public safety rhetoric.
Still, their Washington partnership has produced tangible movement. The voluntary review framework under discussion would let agencies like those overseen by Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick evaluate models for national security and safety risks. Altman plans to preview OpenAI’s newest system during his meetings, a model reportedly capable of sophisticated real-world tasks. The timing feels deliberate. An August 1 deadline looms. Policymakers want concrete proposals. The labs want clarity before they ship even more powerful technology.
Friction remains. Silicon Valley voices accuse the pair of playing a double game. They praise open source in public while quietly pressing for restrictions that would hobble it. Reddit threads and X commentary seized on reports that the companies lobby to limit open-weight releases on cybersecurity grounds even as Altman claims support for openness. One widely shared post asked why humanity must beg permission to own intelligence.
Yet the coordination continues. It spans federal agencies, state legislatures and private conversations with defense officials. Both companies have hired former government staffers and seasoned lobbyists to navigate the maze. Their combined war chest now exceeds what many established industries spend in similar periods. The bet seems clear. Shape the rules early or watch others write them.
Altman returns to the capital at a fluid moment. The Trump administration has signaled openness to industry input but also wariness of ceding too much ground to a handful of labs. Bipartisan lawmakers want oversight without stifling innovation. And foreign competitors, particularly in China, add urgency to domestic safety talks. In that mix the quiet alliance between OpenAI and Anthropic stands out. It may not last as competition intensifies. For now it gives both a stronger voice in the room where rules take shape.
Whether the voluntary framework becomes law or simply a temporary bridge remains uncertain. What is clear is that the two labs have decided the future of AI cannot be left solely to markets or to Congress. They intend to help write the next chapter. Directly.
WebProNews is an iEntry Publication