Sponsors will need to ensure they are complying and submitting full, understandable dossiers if they are hoping to benefit from the US Food and Drug Administration’s (FDA’s) new generative artificial intelligence (genAI) submission assessment.
Although the FDA initially set an ambitious roll-out target of the end of this month in May 2025, the agency moved faster than expected, officially adopting the genAI system on 2 June.
The new AI-assisted scientific review technology, named Elsa, will allow FDA scientists and subject-matter experts to spend less time on tedious, repetitive tasks that often slow down the review process.
According to the agency, the technology is run within a high-security GovCloud environment, and models are not trained on the submitted data, safeguarding the sensitive research and data handled by FDA staff. Elsa is also being used to accelerate clinical protocol reviews, shorten the time needed for scientific evaluations, and identify high-priority inspection targets.
Pierre-Louis Durel, corporate development vice-president at Yseop, a software for AI-powered regulatory writing, spoke to the Clinical Trials Arena about how sponsors need to ensure they are submitting suitable dossiers if they want to reduce submission times by using the FDA’s software.
Abigail Beaney (AB): Could you explain what the FDA’s genAI technology is and how it will impact sponsors and biotechs?
Pierre-Louis Durel (PD): We know that the FDA and the EMA are both adopting AI to accelerate the assessment of the dossier they receive. The agencies are really into accelerating assessment report times to be able to summarise the dossier they receive, so they can identify discrepancies in the dossier more quickly. Agencies are asking pharma companies to provide structured and complete documents so that they can be reviewed by the authorities through genAI.
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalDataCompanies like Yseop are also using AI to help companies complete their dossier, which can reduce sponsor timelines from months to weeks. We also ensure these are structured in a way that agencies will want for genAI review, and, as a result, using AI for the dossier and the approval application is very complementary.
AB: Aside from structural changes to the dossier, what are the other changes sponsors and biotechs need to consider?
PD: What we have seen by working with sponsors is that change management around AI is very important. Yseop is working with the medical writers and bringing in AI to support them with their daily work and make sure their report will help them be compliant with the genAI, as well as helping the writers benefit from AI. Another important thing is the templates of documents, making sure the contents are very clear. Data flow is also very important, and that includes data from the clinical trial – the EDC and the statistical analysis. Making sure this is streamlined is important to create a table that is easier to analyse and understand.
AB: How important is it for sponsors and biotechs to comply with what the FDA is asking?
PD: AI will help authorities identify discrepancies in the dossier faster, and it’s important to note that when different modules are not completely available, AI will still be able to identify this as a lack of compliance. The automation from the FDA incentivises sponsors to provide clean dossiers immediately to avoid back-and-forth conversations over corrections and amendments. With this automation of the assessment at the end, it will really push the sponsor to raise the bar and provide a clean first version of a dossier so they can get to the discussion they want to have with authority and avoid the first interaction being on complexity issues or discrepancies.
AB: Currently, the FDA is using the technology for new drug applications (NDA) and biologics licence applications (BLA) submissions, but do you think they will integrate genAI into IND [investigational new drug] submissions too?
PD: While NDA is on top of priorities in terms of speed, quality and efficiency, IND is also a place where agencies are looking at complex documents that are updated. Sponsors need to summarise these documents to get approval to start the clinical trial faster, so speed, efficiency and quality are still at stake for IND. While genAI is not as important here as the timelines are a bit longer, accelerating IND will contribute to the global acceleration of clinical development processes.
AB: You mentioned that the EMA is also looking to adopt genAI in their drug approvals. Where is Europe in this process?
PD: The EMA is also in the process of adopting AI and that will require companies to properly structure the dossiers, making sure that they are easy to read so the AI can access them. We could envision that they’ll ask to limit the iteration to a single submission. It’s challenging to pinpoint the differences in detail between the EMA and FDA, but we know that the EMA will likely be following the FDA’s lead.
AB: Will there still need to be some interaction from humans at agencies with submissions?
PD: I don’t think the purpose of this is to fully automate everything. These are very complex documents we’re talking about – new drugs with novel endpoints. There are some endpoints with less content available, so we can expect human interaction there. Humans will continue playing a key role, and what we can expect is for AI to enable them to spend time there on new things, rare diseases and novel points. AI is here to really ease the process – get rid of drafting, basic copy and pasting, data description, data interpretation, and more. We will reach some limits at some points, but this will give humans more time to work on these areas.
AB: Where else do you anticipate that the FDA and EMA may be able to introduce genAI capabilities?
PD: The first step is ensuring that this adoption of AI is seen to be positively impacting the processes and accelerating the submission timelines without compromising quality. If we think a bit further, I can envision real-time submission, an immediate and interactive review of the submitted dossier with health authorities. That is a bit further down the line and will require real discussion with health authorities across the globe in an interactive way. If this happens, it will bring drugs to patients even faster.
