AI in Scientific Publishing: Slower, Worse, and More Expensive

by H. Holden Thorp (Science) ... Rigorous human checking of AI-generated research papers is creating bottlenecks as publishers strive to maintain the integrity of the scientific record. The challenge is requiring even more human effort, making the whole endeavor slower and more expensive.

Recently, the ability of large language models (LLMs) to predict protein structures and to accelerate the discovery of new materials has revolutionized both fields. New AI agents can now carry out many aspects of research design and analysis. But there is a dark side. A recent paper describes how an LLM “still struggles in areas requiring nuanced clinical judgment, experimental reasoning, or deep biological thinking and synthesis.” In addition, some of these agents are more likely than humans to engage in what amounts to research misconduct such as cherry-picking data and manipulating statistical analyses until a desired result is achieved. When the agents then generate papers describing the findings, they are likely to make more errors, including hallucinating references, some of which have already made their way into the literature.

Over time, the models may well improve and correct some of these behaviors, but it was recently argued that as long as LLMs are used in research, these errors will always exist. LLMs work not by seeking truth or through logical reasoning but by making probable connections. And efforts to improve these programs to avoid unwanted behaviors may be all for naught because LLMs tend toward sycophantic behavior that keeps the user engaged.

AI tools don’t catch all errors, and they can also generate false positives, incorrectly flagging genuine human research as AI-generated.

The creation of robust human-curated scientific literature has never been more crucial. The propagation of AI slop in the scientific record and generally on the internet is making science less trustworthy. But in parallel, AI enthusiasts are telling the world that the remedies should all be easy to automate, creating the impression that journals can not only catch more errors but also do it more efficiently and cheaply. That’s not the reality for scientific publishing right now or likely in the future. READ MORE

Excerpt from Science: Monday (July 13, 2026) marks the first of two July deadlines for proposals to a $1.5 billion NSF initiative called X-Labs, which it launched less than 2 months ago.

The initiative sparked a furor when it emerged that NSF was diverting money from other research efforts to fund it. And in a radical shift by an agency best known for funding curiosity-driven research at universities, NSF has emphasized its desire to fund a new cadre of nonacademic players, leaving many academics wondering whether they should even bother to apply.

The agency has also created an incentive structure for X-Labs that is radically different from traditional grants, where the payoff is often limited to published results in prestigious journals. Instead, the X-Labs teams would win more funding—up to $300 million over 6 years—for meeting a series of ambitious milestones for scaling up discoveries into commercially viable technologies capable of spawning entire new industries. The first two topics are interconnected quantum systems and photonics and new instrumentation for sensing and imaging.

The NSF solicitation also requires teams to show they can make quick decisions on such things as hiring, procurement, contracting, and assigning intellectual property rights, a not-too-subtle dig at the ponderous bureaucracy at most academic institutions.

But Mini Das, an applied physicist at the University of Houston, says she doesn’t understand why NSF would want to exclude its core audience from an initiative that will require the best scientific minds. “If you’re looking for breakthroughs, you certainly want to include academics, because novel science eventually turns into new technologies,” says Das, who plans to submit a proposal for new 3D imaging technologies that can accelerate medical diagnoses and treatments as well as spur innovation in other fields.

Pamir Alpay, a materials scientist and provost at the University of Connecticut (UConn), applauds NSF for launching the X-Labs initiative. But he says its initial 14 May announcement was a deterrent to UConn. “They were basically saying we don’t want academic institutions to apply,” he recalls. “So we haven’t pursued it.”

The red light for Alpay and others was the requirement that the principal investigator (PI) and other “senior key personnel” devote 100% of their time to the project once it moves from a 9-month planning period into full operation. That would essentially rule out a major role for faculty members, who have teaching and service obligations in addition to their research.

TIP staff also suggested faculty members essential to the X-Labs’s success might need to take a leave of absence or, conversely, accept a lesser role in the project if they wish to maintain their current university status.

A regular NSF grant doesn’t require the PI to devote all their time to that project, and many faculty members manage multiple federal grants simultaneously.

President Donald Trump’s administration has criticized the high overhead rates that universities receive from the government to reimburse them for the cost of doing federal research, he (Alpay) notes. “But I wonder what a new organization without the existing infrastructure would charge in overhead,” he says. “I wouldn’t be surprised if the rates were much higher.”

Although NSF staffers have urged webinar participants to “think big,” applicants must cram their revolutionary ideas for an X-Lab into an eight-page proposal.

NSF has also left open the possibility of expanding X-Labs to other fields that are priorities for the Trump administration such as biotechnology. READ MORE