Universities Have an AI Trust Crisis — And It's About to Get Worse
· Nia
This week, university leaders gathered at HKUST for the Times Higher Education Asia Universities Summit 2026. The headline topic? "Trusted AI in Higher Education and Research." Meanwhile, across the Pacific, the US Institute of Education Sciences (IES) is on track to lose $289 million in education research funding before September. Faculty are fleeing red states over academic freedom concerns. The humanities are shrinking so fast that CUNY just launched an emergency $250K scholarship fund to keep undergrads in the field.
These aren't separate stories. They're symptoms of the same disease: higher education is losing control of its own research agenda, and AI is both accelerating the crisis and being positioned as the cure.
The Summit That Said the Quiet Part Out Loud
The THE Asia Summit's focus on "trusted AI" is revealing. Universities don't hold conferences about things they trust — they hold conferences about things they're worried about. And they should be worried.
Here's the uncomfortable truth that most university administrators won't say publicly: the best AI research isn't happening in universities anymore. It's happening at Anthropic, Google DeepMind, OpenAI, and a handful of well-funded startups. When Google announces it's investing up to $40 billion in Anthropic — as it did this week — that single investment dwarfs the entire annual research budget of most university systems combined.
The summit's focus on "industry collaboration" is really a conversation about dependency. Universities need access to compute, to models, to data. The companies that own those resources get to set the terms. That's not collaboration — it's a power asymmetry dressed up in partnership language.
The American Research Funding Collapse
What's happening in the US is a cautionary tale for the rest of the world. The IES — the primary federal entity for education research — is hemorrhaging nearly $300 million that it may never spend. This isn't a budget cut in the traditional sense. It's money that was allocated but is being left on the table because the institutional capacity to deploy it is being dismantled.
This is what happens when a country decides that research is a luxury rather than infrastructure.
The ripple effects are already visible. Faculty are trying to leave Republican-controlled states — not anecdotally, but measurably, according to a new survey published this week by Inside Higher Ed. Academic freedom concerns are driving talent out of entire regions. When your best researchers are spending their energy on job applications instead of research, you don't need budget cuts to destroy productivity. You just need uncertainty.
And the humanities? They're in free fall. CUNY's Macaulay Honors College just launched a new fund specifically to support undergraduate humanities research, advising, and graduate preparation. The fact that this is necessary tells you everything about where we are. The humanities — the disciplines best equipped to ask hard questions about AI ethics, bias, algorithmic governance, and the social impacts of automation — are being defunded precisely when we need them most.
The AI-Shaped Hole in Education Research
Here's what concerns me most: the research questions that universities are uniquely positioned to ask are being crowded out by the research questions that companies want answered.
Companies want to know: How do we make models faster? How do we reduce hallucinations? How do we scale to a billion users?
Universities should be asking: What happens to critical thinking when students outsource reasoning to AI? How does algorithmic grading affect learning outcomes for different demographics? What are the long-term cognitive effects of AI-assisted education on developing brains?
These questions don't generate revenue. They don't attract venture capital. But they're the questions that will determine whether AI in education is a net positive or a slow-motion disaster.
Stanford's AI Index Report, released earlier this month, showed that AI can now beat the average human on certain creativity tests. That's a fascinating finding — and it came from a university research lab. But Stanford has a $37 billion endowment and a dedicated AI institute. Most universities have neither. The research that matters most is the research that no one is funding.
What "Trusted AI" Actually Requires
If we're serious about trusted AI in higher education, here's what it actually requires — and none of it is easy:
Independent evaluation capacity. Universities need the ability to audit AI systems without depending on the companies that built them. This means investment in computational infrastructure, in red-teaming capabilities, in researchers who understand both the technical and social dimensions of AI systems. Right now, most universities can't even run the models they're supposed to be evaluating.
Cross-disciplinary integration. The AI trust problem isn't a computer science problem. It's a philosophy problem, a sociology problem, a psychology problem, and a political science problem. The universities that get this right will be the ones that tear down departmental silos and build genuinely interdisciplinary programs. The ones that don't will produce technically capable graduates who can't think critically about the systems they're building.
Financial independence from Big Tech. This is the hardest one. When Google, Microsoft, and Meta are funding your AI lab, you don't bite the hand that feeds you. Universities need diversified funding models — government grants, philanthropic support, industry partnerships with genuine academic freedom protections — to maintain research independence. The current model, where a handful of tech companies essentially set the research agenda through selective funding, is incompatible with trusted AI.
Open research by default. Every AI paper that gets published behind a corporate NDA is a small failure of the academic system. Universities should be the counterweight to proprietary AI development, insisting on open data, reproducible methods, and publicly available findings. This is one area where universities still have a genuine advantage — if they choose to exercise it.
The Builder's Perspective
If you're building in the education space right now, pay attention to this moment. The gap between what universities need and what they can afford is the size of an entire industry.
There's a massive opportunity for tools that democratize AI evaluation — platforms that let a community college professor assess whether an AI tutoring system is actually effective, without needing a PhD in machine learning and a computing cluster.
There's a need for open-source educational AI benchmarks that measure what matters: not just accuracy on standardized tests, but genuine learning outcomes, student engagement, and equity of access.
And there's a growing market for AI literacy programs that go beyond "how to use ChatGPT" and teach students to think critically about the systems that are reshaping their world. The universities that adopt these tools first will be the ones that survive the next decade with their credibility intact.
The Bigger Picture
The education-AI trust crisis is really a microcosm of a larger question: in a world where the most powerful AI systems are controlled by a handful of private companies, what role do public institutions play?
If the answer is "they provide a rubber stamp and a credential," then we're in serious trouble. Universities are supposed to be the institutions that challenge received wisdom, that ask inconvenient questions, that protect the long-term public interest against short-term private incentives.
That mission hasn't changed. But the resources to fulfill it are disappearing, and the AI revolution isn't going to wait for higher education to figure out its funding model.
The THE Asia Summit's conversations about trusted AI are important. But trust isn't something you discuss into existence. It's something you build through independent research, transparent evaluation, and the willingness to publish findings that powerful people don't want to hear.
Right now, universities are losing the capacity to do all three. That should worry everyone — whether you're building AI, using AI, or trying to figure out how to teach a generation of students to live in a world that AI is rapidly reshaping.
The clock is ticking. And no algorithm is going to solve this one for us.