This morning's three AI news items seem to cover the US government, data center financing, and AI safety. But they all address the same core question:
When AI is no longer just a chat tool, the real challenges become 'who can deploy it, who can afford the computing power, and who decides it's safe enough to go live.'
① US Government Website Used Qwen, Removed Quickly After Exposure
According to Reuters, the US government's Federal Register website once offered an AI search option powered by Alibaba's Qwen model, allowing users to search publicly available federal regulations and public comments.
Notably, just a week prior, the US FBI publicly accused Alibaba and other Chinese AI companies of large-scale distillation using US models. Whether Alibaba engaged in improper copying is currently an allegation and not a legally confirmed fact.
The Qwen search function was removed soon after social media attention. The US National Archives and the White House did not respond to Reuters' inquiries.
However, this issue shouldn’t be simplified to "using a Chinese model equals government data leakage." The Federal Register handles data that is already public, and Qwen is an open-weight model that can be downloaded and deployed on one’s own infrastructure.
The real question is:
Where is the model running? Did the data leave the government’s secure boundaries?
The same AI model poses a completely different risk when downloaded and run on a secure server versus sending data to a model provider’s cloud API.
② Crusoe Raises Another $3.9 Billion, Valued at $30.9 Billion
The second story reveals the financing behind AI.
AI infrastructure company Crusoe announced a preliminary close of its Series F funding round, raising $3.9 billion and reaching a post-money valuation of $30.9 billion.
Crusoe states that its platform’s total contracted value exceeds $140 billion, with data center and cloud capacity contracts over 6 GW, about 1 GW of which is currently operational.
The company originally started by using surplus energy for crypto mining but has gradually evolved into a one-stop infrastructure provider offering power, data centers, GPU infrastructure, cloud services, and even managed inference for major AI clients.
And this is not an isolated case.
In early September, SasaDaily reported that Crusoe was said to have secured a $13 billion, five-year AI cloud contract with Jane Street. Less than two weeks later, they closed this $3.9 billion financing.
The AI boom is creating a new industry previously unseen: Neocloud.
These companies may not train the top models themselves but specialize in building the massive GPU, power, network, and data center capacity needed by those who do.
③ Amazon Joins AI Safety Debate, But Doesn’t Support Industry-wide Slowdown
The third story is one of Silicon Valley's most sensitive issues: should AI development slow down?
Amazon told Reuters that AI models should only be released when they are "ready and safe,” which requires rigorous testing and strong safeguards.
But Amazon did not back the recent calls by some AI leaders for an industry-wide slowdown to simultaneously reduce AI development speed.
This distinction is critical.
One approach is:
Continue fast AI development, but require each company to prove its models have undergone sufficient testing.
The other:
If capability grows too fast, should leading AI labs coordinate to slow development together?
Amazon currently leans toward the first approach and suggests that both industry and government should work together to establish proper protections.
What’s Truly Worth Watching Today: AI Competition Moves into Operational Reality
Looking at these three stories together shows that the next phase of AI competition is no longer just about benchmarks.
- Models: Not about whether you can download them, but where they are deployed, data boundary policies, and supply chain acceptance.
- Computing Power: Stronger models require exponentially more data centers, electricity, GPUs, and capital.
- Safety: Higher capabilities force companies to answer the question of what conditions must be met to release a model.
Previously, the question was: "Which AI is the strongest?"
Going forward, enterprises and governments will ask more often:
Can this AI be deployed? Can data be shared with it? Can the computing cost be afforded? Who is responsible if something goes wrong?
This is where AI truly becomes more difficult once it moves from demo to the real world.
Today, let's advance alongside AI a bit:
Learn one AI skill daily.
Save a little more time every day.
Improve your capabilities step by step.
SasaDaily, growing with you.
