AI Giants Chafing Under Washington’s Unexpected Control
POLICY WIRE — Washington, D.C. — Imagine having developed a brainier-than-human artificial intelligence, something potentially revolutionary, and then having to wait for bureaucrats in Washington to...
POLICY WIRE — Washington, D.C. — Imagine having developed a brainier-than-human artificial intelligence, something potentially revolutionary, and then having to wait for bureaucrats in Washington to sign off on its public release. That’s the peculiar, sometimes baffling, reality now confronting America’s leading AI firms, OpenAI — and Anthropic. What was once the wild west of innovation is becoming — to Silicon Valley’s chagrin — increasingly domesticated.
It’s not just a cybersecurity audit; it’s a full-blown federal bottleneck, spearheaded by the Trump administration. Earlier this week, OpenAI, the crew behind ChatGPT, found its new model, GPT-5.6 Sol, wouldn’t be hitting the open market just yet. Instead, access was cordoned off, exclusively for a handful of clients deemed acceptable by the White House. This wasn’t their idea. Hours later, competitor Anthropic — the folks responsible for the Claude chatbot — got a similar nod for its formidable Mythos 5, a restricted clearance that allows a limited run for some trusted partners. But Anthropic already knew what it was like to be grounded. Just weeks prior, it had to pull two models offline entirely, Mythos 5 included, because Uncle Sam said so. [QUOTE_PLACEHOLDER]
This whole government-says-go-or-stop charade comes after Anthropic earlier floated a rather uncomfortable truth about its Mythos model: it’s wicked good at sniffing out software vulnerabilities. So good, in fact, that it could be weaponized. David Sacks, an investor close to Trump and a key advisor, laid it out on a recent podcast, noting that Anthropic CEO Dario Amodei effectively got Washington all sorts of twitchy. Sacks said: Dario came to Washington a few months ago, back in April, and basically said that he had created a cyber weapon called Mythos. And he spiked the cortisol level, got everyone really worried. He then conceded there was some truth to it in terms of the sense that this model had advanced cyber capabilities. And just like that, the administration’s hawk-eye settled squarely on advanced AI. Never mind the niceties; this is about national security.
Because, make no mistake, it isn’t simply an informal request. An executive order, signed by Trump in June, established a framework. It basically grants the government a 30-day window to scrutinize the big AI systems before they can be released to the masses. Voluntary, they say, but really? Is anything truly voluntary when the White House is involved? Critics, such as U.S. Rep. Lori Trahan, a Massachusetts Democrat, aren’t mincing words. Trahan stated: the Trump administration is deciding company by company who gets access to the newest AI model. No law. No process. No oversight. Just appointees in Washington deciding who’s in — and who’s out.
OpenAI, for its part, insists its Sol model is actually benign, proclaiming it’s better at helping people find and fix vulnerabilities than at launching cyberattacks. Still, it admits to uncertainty— especially if its model combines with other tools. That uncertainty, along with the model’s broader step change in capabilities, is why we’re pairing the model’s increased capabilities with stronger safeguards and a phased release, the company declared. Meanwhile, the most contentious aspect might be less about cyber war — and more about market intervention.
The Pentagon already put Anthropic on a naughty list, calling it a national security risk over ethical concerns with AI in warfare. Then Trump himself reportedly told federal agencies to cut ties with Anthropic’s Claude. The company slapped back with a lawsuit, which is currently slugging it out in federal courts. So you can see it’s not exactly a partnership made in heaven. Yet, even with all this friction, Anthropic later voiced that it was pleased by the partial release of Mythos and will continue to work with the government to expand access. One can only imagine the conversations unfolding behind closed doors, no doubt involving some significant horse-trading.
But the government’s muscle-flexing carries geopolitical weight beyond Silicon Valley’s squabbles. If America’s cutting-edge AI innovation is effectively shackled by regulatory whims, it impacts a global tech ecosystem that often looks to the U.S. for leadership. Nations in South Asia, for instance — think Pakistan, a country consistently battling cyber threats and looking to leverage AI for everything from agriculture to national defense — rely on the open, rapid deployment of advanced tools. When models are held back, it creates a ripple effect. Competitors like China, unencumbered by similar transparency demands or the thorny U.S. voluntary compliance framework, could accelerate their own deployment and adoption rates, potentially creating a significant advantage. Alex Stamos, a cybersecurity expert from Stanford, wasn’t shy, declaring: If the administration is honest about wanting the United States to beat China in this race, then this is about the dumbest thing they could possibly do.
These latest developments also come as both OpenAI — and Anthropic are eyeing splashy Wall Street debuts. SpaceX’s record-setting June 12 initial public offering undoubtedly put stars in their eyes. But how do you pitch an IPO when the product you’re selling could, at any moment, be put in a government sandbox? Trump has even mused about the government owning stakes in these companies, with pieces could be given to the American public, where the American public essentially becomes a partner with the companies. It’s a curious concept, for sure.
What This Means
This isn’t just about cybersecurity; it’s about control— raw, unadulterated political and economic control over what might just be the most transformative technology of our time. For Washington, the implications are vast: demonstrating sovereign power over tech, potentially shaping an industry in its own image, and preventing (or at least delaying) future systemic shocks. But it’s a tightrope walk. Too much intervention risks stifling innovation and driving talent—and potentially, the future of AI itself—to less regulated shores. Think places eager to embrace unburdened tech development, a scenario China would undoubtedly capitalize on. For the AI industry, it’s a stark choice: bend to the federal will, even when it feels arbitrary, or face an existential battle against an increasingly assertive government. This heavy-handed regulatory posture could reshape the global competitive landscape, granting an edge to nations that adopt a more permissive, or even strategically state-directed, approach to AI development. Companies are caught between market demands — and a federal hammer. And frankly, this might just be the opening salvo in a much longer, uglier fight over who truly calls the shots in the AI revolution.
