- The Deep View
- Posts
- Why AI took early blame for the FAA flight outage
Why AI took early blame for the FAA flight outage

Welcome back. The IMF is warning that circular financing could turn the industry’s financial risks into problems for the broader economy. Meanwhile, OpenAI is embedding four independent auditors inside its operations. That’s a welcome step, but the real test will be how much of what they find the public gets to see. And while AI quickly became a suspect in the FAA flight outage on Friday, the fact that it turned out to not be the culprit doesn't lessen the fact that Congress and the public have every right to be concerned about the impact of the agency's new AI program on one of the nation's most critical systems. —Jason Hiner
1. Why AI took early blame for the FAA flight outage
2. Why the IMF is worried about AI circular financing
3. The real test of OpenAI’s audits is disclosure
GOVERNANCE
Why AI became a suspect in the FAA outage
One of the consequences of AI being everywhere is that it also becomes one of the first things to blame when stuff goes sideways.
Friday's FAA systems outage in the US led to the first "ground stop" since January 2023 and resulted in tens of thousands of flights being cancelled or delayed and over 1 million travelers being stranded. Speculation quickly escalated that it could be AI-related. However, on Friday, Homeland Security secretary Markwayne Mullin put the kibosh on those rumors saying the department found no evidence that AI or cyberattacks were the root cause of the outage, while reporting that an investigation would be on-going.
In a briefing on Friday, Transportation secretary Sean Duffy said, "We started to have inconsistency in the data from the radars and our communication with airplanes, and we have procedures for this." He stated that newly installed digital equipment continued to operate normally during the outage while it was the older analog systems that malfunctioned and led to the issues.
The outage ended up lasting almost five hours from 3:30 p.m. to 8:15 p.m. ET on Friday and Duffy ultimately blamed the outage on a "contractor error or a faulty piece of equipment" as part of a networking upgrade from traditional copper wires to fiber optic cables.
However, the AI speculation emerged because the FAA is in the midst of implementing an $875 million AI-powered system called SMART (Strategic Management of Airspace, Routes and Trajectories). The new platform began rolling out in September and the agency described it as a "cutting-edge tool that centralizes 200 data streams—including weather patterns, flight paths, traffic flow, and controller staffing metrics—into one platform."
During early testing in the Washington, D.C. area, it helped reroute 180 flights and save 10 minutes 80 gallons of fuel per flight by diverting planes around weather and other conditions. SMART was expanded nationwide on November 10, three days before ground stop.
Last month, members of the the House Committee on Transportation and Infrastructure requested an audit of SMART to insure that the "proper safety guardrails" were in place for implementing AI into such important software and specifically called out concerns about the potential impact AI hallucinations could have on mission critical systems.

AI is clearly becoming part of everything, and that now includes conversations about risks to societal-level systems. It's certainly understandable that members of Congress would want assurance that the FAA has evaluated and understood all of the potential risks in integrating AI into such critical infrastructure such as the nationwide flight navigation systems. We saw on Friday how the smallest errors can lead to problems with cascading effects that can impact the lives of so many people. And that doesn't even bring up the life-threatening consequences naturally involved in every single airline flight. Putting any of that in the hands of unpredictable, non-deterministic systems is a scary thought, and it's only been amplified by the surprising failures we've learned about from the Hugging Face incident and the myriad of similar occurrences like it that have been revealed in the months since.
TOGETHER WITH AMD
Are Your AI Agents Costing More Than They Should?
As organizations deploy more AI agents, cloud inference costs can add up with every task. But running AI locally could change the economics. The AMD Tokenomics Calculator helps you compare local, cloud, and hybrid deployments to uncover potential savings and optimize AI spending.
Understand your AI costs. Explore how infrastructure requirements and token usage impact your total AI spending.
Compare local, cloud, and hybrid deployments to identify opportunities to reduce recurring inference costs.
Get more from your AI investment. High-performance agentic PCs featuring AMD processors can run multiple AI agents locally, completing tasks up to 6x faster while reducing reliance on cloud inference.
MARKETS
Why the IMF is worried about AI circular financing
Circular financing, in which investors put money into companies that in turn hand it back for buying products or other investments, may appear to offer quick returns, but the International Monetary Fund (IMF) warns the practice carries significant long-term risks.
The IMF estimates that circular financing has grown to just over $1 trillion, with two-thirds funded by debt, said IMF director Kristalina Georgieva in an interview with CNBC. While that is not a "huge size," she does say it represents a rapid increase. This is especially notable as the world economy relies on AI, and any fluctuation in that market can affect worldwide economic growth.
“We have a world economy that is standing on one leg on the AI leg. If this leg wobbles, that can lead to financial stability troubles,” said Georgieva in the interview.
The debate over circular financing is especially timely now that Nvidia, one of the practice’s leading proponents, is now in the spotlight. A scandal over its CEO and co-founder’s extramarital affairs forced him out and rippled through the stock market, wiping more than $1 trillion off the company’s market value over the past month. Beyond shaking investor confidence, the upheaval raises the question of what happens if Nvidia’s next leader pulls back on these creative financing deals.
That now looks likely: according to The Wall Street Journal, CFO and interim CEO Colette Kress was skeptical of many of the deals when they were signed, and Huang himself had begun laying the groundwork to scale back the approach before he resigned.
Nvidia’s own disclosures show how far this practice has spread. As of July, its equity investments were valued at $99 billion, up from about $7 billion a year earlier and about $2.2 billion two years before that, CNBC reports. Most of the money has gone to frontier AI labs, with Kress saying Nvidia has invested nearly $50 billion in them. However, the companies likely to feel the impact of any Nvidia investment pullback are the neoclouds Nvidia has invested in, which have more to lose, Shiva Rajgopal, Ph.D., professor at Columbia Business School, told The Deep View.
"The first to feel it would be the smaller AI cloud providers and model developers that rely on Nvidia for capital or capacity guarantees, not the big cloud companies, which fund themselves more or less," said Rajgopal.
Huang reportedly even warned its neocloud partners about it at its conference in Berlin last month, telling them that they would need to secure upfront payments from customers so it was easier to finance and Nvidia wasn't the sole financer. The same WSJ report cites an analyst note which predicts that Nvidia's financing pullback could take $50 billion or more out of the market.
This could, in turn, affect Nvidia itself. "Accounting history is a caution here: in the late-1990s telecom boom, equipment makers financed their own customers, and revenue fell sharply when the financing stopped. Nvidia’s situation is different, with far stronger customers, but that’s the question investors will ask," added Rajgopal.

The turmoil surrounding Nvidia goes far beyond interest in Huang's personal life; it's more about what a change of guard means for leadership ideology, which significantly impacts the entire AI industry and can even have repercussions for the world's economy, as Georgieva warns. However, this is still a separate question from what will happen to AI itself, and that answer is clearer: the technology is not going anywhere. While Sam Altman has some self-interest in sharing optimistic statements about the industry, the logic of what he said at Semafor’s Silicon Valley & The World conference on Monday is sound: "The revenues and demand are growing so strongly that I think the industry will be fine."
TOGETHER WITH FINGERPRINT
Grok Bot went quiet. Here’s how to spot it anyway
xAI's Grok Bot signs in to websites, works through inboxes, and finishes multistep tasks for real users.
Until recently, it announced itself with signed requests. Around October 9, that stopped.
With signatures gone, it works from a real cloud browser with no verifiable identity, no automation flags, and no telltale IP address, so most bot defenses see an ordinary visitor.
Fingerprint's Bot Detection Smart Signal now labels it by name, so you know who is visiting. Read how to detect Grok Bot with Fingerprint.
GOVERNANCE
The real test of OpenAI’s audits is disclosure
OpenAI is opening its doors to independent auditors as the stakes to develop safe frontier AI get higher.
On Monday, the AI lab announced that it has signed agreements with four third-party auditors: Apollo Research, METR, Transluce, and Goodfire. These assessors will be embedded across OpenAI's operations, allowing them access to the same internal resources that its researchers have, including model checkpoints, source code, data, monitoring systems and technical documentation.
The announcement marks the first phase of OpenAI's embedded assessments, the company said, with initial evaluations focusing specifically on alignment and safeguards in training and internal use.
Each of these four assessors are responsible for evaluating different parts of OpenAI:
Apollo Research will assess whether models can covertly work together against a user's intentions, and if OpenAI's existing safeguards can detect and mitigate this behavior.
Goodfire's interpretability researchers are primarily responsible for training run monitoring, debugging, and alignment.
METR will be responsible for two assessments, one focused on spotting misaligned behavior within OpenAI's "largest reinforcement learning run to date," and the second being a "stress testing exercise" involving red teaming.
Transluce will conduct post-training alignment evaluation on selected public and unreleased OpenAI models.
OpenAI also developed a six-part protocol for third-party auditors, aiming to create a specific scope, secure environment and proper oversight, as well as protocols around public disclosures. Following the initial assessments, OpenAI will publicly report some of the findings and iterate on them for long-term, consistent embedded evaluations. However, the AI lab said that it retains "the right to redact confidential information," as well as request "responsible disclosure delays" to give OpenAI time to act on the findings.
"Especially in the early stages of embedded assessments, we expect technical access challenges, space constraints, evolving security protocols, and other unexpected issues to arise," OpenAI noted in its blog post.
This decision follows a similar move by rival Anthropic, which announced in October that it would give "employee-like access," including access to hardware, office space, and internal tools, to six assessors from METR to evaluate the risks of its frontier models. It also follows OpenAI, Anthropic and several other frontier labs signing a "morally binding" White House pledge to self-regulate, an accord which included promises for partnerships with external auditors.

Third-party evaluators have already proven themselves to be a critical part of holding AI labs accountable for building safe frontier systems. Last month, METR published findings that a three-week evaluation of OpenAI's systems, which coincidentally took place during the Hugging Face incident, uncovered a number of vulnerabilities, including ways for agents to circumvent and conceal actions from human monitoring. However, these third-party evaluations only work when the auditors performing them are not muzzled, such as in the case of a recent Reuters report claiming that OpenAI and Anthropic had attempted to control what evaluators published to prevent IP leakage. Independent audits with restrictions on what can actually be publicly released in some respects defeat the purpose, as they threaten the power of accountability that these assessors have. Though OpenAI is acting responsibly in embedding these evaluators at all, the question still remains what, exactly, the public will be able to learn from them.
LINKS

Calif. AG Rob Bonta sues OpenAI for rogue agents
Jensen Huang allies lobby Nvidia board to reinstate him as CEO
New rule requires data centers in five countries to license US AI chips
Apple launches study to explore AI use for health
Google warns agentic AI threat activity has crossed a significant threshold
Knit Health raises $35 million for AI-driven patient care prediction

Qwen4-Max: Alibaba's new flagship 2.4T-parameter MoE model with 1M-token context
ChatGPT Health: will soon work with Mac app, Dots, more, according to report
Ask Maps: updates make it easier to plan trips, book errands, and prepare for traffic
Cohere: Launched Video Intelligence in Compass

A QUICK POLL BEFORE YOU GO
Do you think AI labs need to be more transparent about their third-party auditing? |
The Deep View is written by Nat Rubio-Licht, Sabrina Ortiz, Jason Hiner, Faris Kojok and The Deep View crew. Please reply with any feedback.

Thanks for reading today’s edition of The Deep View! We’ll see you in the next one.

“The shadows appear more accurate in [this image].”
|
“Reflections don’t match in [this image].”
|


If you want to get in front of an audience of 750,000+ developers, business leaders and tech enthusiasts, get in touch with us here.













