Grant & Growth #2 - NIH Warns Off Overuse of AI in Grant Writing

The limit becomes effective from September 25, 2025, and new, renewal, resubmission, or revision applications from an individual Principal Investigator/Program Director or Multiple Principal Investigator for all council rounds in a calendar year count towards the allowance.

In addition, NIH will employ technology to detect AI-generated content and identify non-compliance with funding guidelines.

The way the note is written suggests that NIH observed a break in the data perhaps explaining the threshold of 6 per year. Shocking is that apparently some applicants have recently submitted 40 grants in a single cycle 👀, which strikes me as pretty clear prima facie evidence of AI churn. I am surprised that any institution permitted this or that any investigator thought it was a good idea.

The new application limits are expected to affect only a small number of PIs. The allowance, which excludes R13 conference grants and training grants (T series), seems ample for almost any scientists, even one with multiple research grants.

In my opinion the limit on applications is a clever and strategic way to shift the conversation around AI in grants. It removes the major incentive of over-reliance on AI, which is the ability to generate large number of proposals with little effort. Combined with the surveillance, and the risk of being referred for research misconduct if you submit an application substantially developed by AI, this is likely to keep AI where it should be - another tool for scientists to create great proposals, but not a replacement of the human at the center of the proposal.

Here are just a few that resonate with me from a recent post by iPEC Coaching:

What do you think?

A newsletter I recently subscribed to - Your Local Epidemiologist - published a post on the 23rd about what is going on at NIH. Behind the loud headlines of grants being cancelled, sometimes reinstated and then cancelled again and the fights with Columbia, Harvard and others, there is a quiet dismantling.

In the data they share (from Jeremy M. Berg via GitHub) you can see how far behind last year the NIH spending is this year. Some of the difference will be grants canceled because they ‘no longer align with the administration’s priorities’, which has taken the shape of eliminating entire program such as the Fogarty International Center, targeting topics like the health of minorities, gender studies and vaccine confidence. But not all the difference.

Stark here is the status of T mechanisms, a group dominated by the T32 training grants that support graduate students, residents, fellows and post-docs. I am sure many of them will get funded before the end of September, making for a highly compressed late summer for the program and grants management teams at NIH - my heart goes out to my old team at the Center for Cancer Training at NCI. But why the extreme lag?

I suspect that some of the discrepancy at NIH is due to intentional slow down of the complex mechanisms of how grants are actually disbursed. The financial systems at NIH (and probably much of the government) are antiquated and complex, and as a result largely opaque. Program teams have very limited view of the financials, and since they are facing the research community often find themselves unable to explain what is going on. I suspect that the infiltration of NIH and HHS by DOGE has a lot to do with it, and that they have their hands on levers deep in the financial systems that most people don’t see.

The harm is widespread - the Center for American Progress has mapped out overall federal grant terminations. In their Fig. 1 they display the information per student, and this is interesting, politically, as there is no clear relationship to voting patterns. Hardest hit is South Dakota, next is Hawaii…

Ms. Jetelina and Dr. Marnik of the Your Local Epidemiologist newsletter point to some things you can do which are great. I would agree and say reaching out to your elected representatives is a good step. Turns out that most people in America want more research and that includes people from both sides of the aisle.

Also on the 23rd of this month, NCI posted the Funding Policy through end of September, when the government fiscal year ends. Recognizing that NIH overall is navigating extremely turbulent waters, NCI has taken an approach aimed at sustaining as many scientists as possible. However, the news is not good, and it could be a harbinger of worse to come.

One of the sources of the extreme pressure was the unexpected edict half-way through the fiscal year, to fund around half of the remaining R01 grants in their entirety. Let me explain.

If, for simplicities sake, we imagine a 5-year R01 costs $200 a year then under the traditional funding approach NCI would have allocated $200 in the year the grant started. Then the next year it would have to have another $200 and so on. If in contrast it has to fund the entire grant, then it would cost $1,000 in the first year.

While full-year funding is desirable - and would make budgets much easier to manage with fluctuating Congressional allocations - the transition from a year-by-year to a full funding approach is very challenging without additional funds.

Again, a simple example. Let’s assume NCI awards 500 new R01 grants in a given year, each costing $200 for that first year it would spend $100,000. But if half of these cost an additional $800 because they are full-year funded, these alone would consume $250,000. Now the new policy started mid-year, so we might say it affected only 100 new R01 grants - but these still would cost the same as the 500 originally budgeted for. The only way to make the math work is to reduce the number of grants, which explains the collapse of the payline to the 4th percentile.

If the near-40% cut in funding for NIH that has been requested by the White House for fiscal 2026 guides Congressional action, then we may be looking back at these paylines with nostalgia. The results for science would be catastrophic. Here’s to hoping it doesn’t come to that.