New investigation: US repeating cloud-era mistakes as it rushes into AI
A ProPublica investigation published April 6 by Renee Dudley warns that the federal government is moving into artificial intelligence the same way it did into cloud computing a decade ago—fast, cheap, and with the same weak oversight and vendor dependencies that left agencies exposed.
The White House has framed AI as a national competitiveness priority and opened government access to commercial models at cut-rate prices: OpenAI’s ChatGPT for about $1 per user, Google’s Gemini for $0.47, and xAI’s Grok for $0.42. Dudley argues this mirrors the early 2010s push for cloud adoption under the Obama administration, when rapid procurement and attractive “free” offers led to long-term lock-in and security trade-offs.
Three lessons from the cloud transition
- “Free” upgrades can be lock-in mechanisms. Microsoft’s 2021 pledge of $150 million in security services to the federal government turned into a practical dependency: agencies who accepted the upgrades faced high switching costs later. Even Microsoft and OpenAI have since disputed terms of their own AI partnership—an indicator of how fraught big-tech AI contracts can be.
- Oversight needs funding and muscle. FedRAMP (the Federal Risk and Authorization Management Program), created in 2011 to vet cloud services, was pressured to approve major products despite cybersecurity concerns. ProPublica reports the program now operates “with an absolute minimum of support staff” and “limited customer service,” while a GSA spokesperson defended it as operating “with strengthened oversight and accountability mechanisms.” Former staffers say it has at times functioned like a rubber stamp.
- Independent audits aren’t truly independent if vendors pay them. As FedRAMP’s in-house capacity shrank, third-party auditing firms hired and paid by cloud providers picked up vetting duties. Understaffed agencies then relied on those vendor-funded ratings rather than conducting deep, independent reviews.
The immediate risks for AI adoption
The General Services Administration (GSA) has warned that AI “usage costs can grow quickly without proper monitoring and management controls,” and has urged agencies to set usage limits and monitor consumption. But Dudley’s reporting highlights deeper structural problems: underfunded oversight bodies, a vetting ecosystem financed by the very vendors being reviewed, and agencies that lose leverage once a technology is embedded.
Dudley’s closing warning is blunt: “The implications of this downsizing for federal cybersecurity are far-reaching.” For crypto and AI advocates watching government adoption closely, the takeaway is clear—cheap, rapid access to powerful models can produce long-term security, fiscal, and governance liabilities unless procurement rules, funding for independent oversight, and anti–vendor-lock-in safeguards are strengthened.
Read more AI-generated news on: undefined/news
