iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
The FDA uses an internal generative AI assistant called Elsa to help employees with regulatory and scientific work. The agency has described efficiency gains, but public evidence does not show that Elsa independently approves drugs or that it has shortened approval timelines. CNN reported in July 2025 that FDA staff alleged Elsa sometimes fabricated studies or misrepresented research; those accounts are allegations, not a published error-rate study. The FDA’s May 2026 update describes Elsa 4.0 and human expert verification, but does not quantify accuracy or speed.
Is the FDA using AI to approve drugs?
No public source cited here establishes that Elsa makes drug-approval decisions on its own. The FDA describes it as an assistant for agency employees, including scientific reviewers and investigators. A reviewer may use AI-generated summaries or other assistance as part of their work, but the final scientific judgment remains a human responsibility.
That distinction matters: using AI to help with work connected to drug review is not the same as delegating the decision to approve a drug. The FDA’s June 2, 2025 launch announcement presents Elsa as an internal tool for employee tasks, not as an autonomous approval system.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWhat is Elsa, and what does it do?
Elsa is the FDA’s internal, large-language-model-powered generative AI assistant. The agency says it can help staff read, write, summarize, and work with information. Its described applications include preparing adverse-event summaries, comparing product labels, generating code, reviewing clinical protocols, supporting scientific evaluations, and helping target inspections. These are FDA-described uses; they are not evidence that every task is automated or that each use has been independently validated.
#1 Best Overall
When the FDA launched Elsa agency-wide in June 2025, it said the tool ran in a high-security GovCloud environment and that its models did not train on regulated-industry submissions. Those are agency statements about its deployment and data handling, not an independent security assessment.
What did the FDA say about speeding up reviews?
In a May 8, 2025 announcement, the FDA said it had completed its first AI-assisted scientific-review pilot. Deputy Director Jinzhong (Jin) Liu of the FDA’s Office of Drug Evaluation Sciences, Center for Drug Evaluation and Research, said: “This is a game-changer technology that has enabled me to perform scientific review tasks in minutes that used to take three days.”
That is a named official’s example of certain tasks during a pilot—not a controlled estimate of average time saved, a result for a complete drug application, or proof that an approval happened faster. The same announcement set a June 30, 2025 target for agency-wide integration; the FDA then announced Elsa’s launch on June 2.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →The agency’s May 6, 2026 update describes Elsa 4.0 and a broader integration with HALO, a platform that the FDA says consolidated more than 40 application and submission data sources, systems, and portals. That figure describes the scale of the integration, not Elsa’s accuracy or its effect on review duration. The 2026 FDA announcement describes expanded functionality and says subject-matter experts verify inputs, analytical processes, and how outputs are implemented. These remain agency descriptions; the release does not independently establish that earlier concerns have been resolved.
Rank #3
Did Elsa make up studies or misrepresent research?
CNN reported on July 23, 2025 that current and former FDA employees alleged Elsa had fabricated nonexistent studies and misrepresented research. CNN’s account relied in part on anonymous staff sources and documents. One unnamed employee was quoted as saying, “It hallucinates confidently.” These are reported staff allegations, not findings from a published audit, and they do not establish how often errors occurred or whether all uses of Elsa were unreliable.
In a CNN broadcast transcript, FDA Commissioner Marty Makary described the reviewer’s role this way: “And so it’s the responsibility of the scientific reviewer to click on that link that ELSA provides and look at the study and read the abstract.” That statement explains the agency’s stated expectation of human checking; it does not show that every citation or link Elsa produced was reliable.
Early feedback was not uniformly negative. A June 4, 2025 report by RAPS described officials’ concerns about outdated answers as well as one reviewer’s positive experience querying reports and checking Elsa’s responses. Taken together, the accounts point to a tool whose usefulness and reliability may depend on the task and on verification, rather than proving either universal failure or dependable accuracy.
What is known—and not known—about Elsa’s performance?
The public claims and evidence available in the cited sources are different kinds of information, so they should not be treated as equivalent:
Best Value
- Used Book in Good Condition
| Evidence | What it supports | What it does not establish |
|---|---|---|
| FDA’s stated use cases and deployment descriptions | The agency intends Elsa to assist employees with information and workflow tasks. | Independent proof of accuracy, security, or faster drug approvals. |
| Liu’s pilot anecdote, published by FDA in May 2025 | One official said certain scientific-review tasks took minutes instead of three days for him. | An average time saving, a controlled comparison, or a faster complete review. |
| Staff accounts reported by CNN and RAPS | Some employees reportedly encountered fabricated or misrepresented research, outdated answers, or found Elsa useful when they checked its responses. | A verified error rate, the frequency of errors, or a conclusion that every Elsa use succeeds or fails. |
| FDA’s Elsa 4.0 and HALO update in May 2026 | The agency says the system was expanded and integrated with more than 40 data sources, systems, and portals, with expert verification. | Independent confirmation that earlier issues are fixed or that reviews or approvals are faster. |
The cited public materials do not provide an independently measured Elsa hallucination or error rate, nor an independent quantified estimate of its effect on drug-approval timelines. Without those measures, it is not possible to determine from these sources how often Elsa makes mistakes or whether it has shortened approvals.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How should readers interpret the “speed up” and “horrible mistakes” claims?
- “Speed up”: The FDA has promoted Elsa as a way to reduce time spent on repetitive work, and one official described a substantial time saving on certain tasks. That is not the same as demonstrating faster drug approvals.
- “Horrible mistakes”: CNN reported staff allegations of fabricated studies and misrepresented research. The reporting warrants attention, but it is not a published measurement of error frequency.
- Human oversight: The FDA says reviewers should verify information and, in 2026, described subject-matter experts as checking inputs, analyses, and implementation of outputs. Assistance still requires a person to assess whether a response is accurate and relevant.
- Changing system: Elsa 4.0 and the HALO integration are later developments than the 2025 rollout. The FDA’s update describes those changes but does not publish an independent evaluation showing how they affected accuracy or review times.
For further detail, the primary sources are the FDA’s launch announcement, its pilot announcement, and its Elsa 4.0 update. CNN’s reporting and broadcast discussion are available in its article reproduction, News Central transcript, and broadcast transcript.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
