Is CRISPR Safe? Side Effects, Risks, and What the Science Shows

"Is CRISPR safe?" is the single most common question people ask when they first learn about gene editing. It is also one of the hardest to answer honestly, because the truthful response is neither a simple yes nor a simple no. It is: "It depends on what you mean by safe, what disease you are treating, and how you weigh manageable risks against the alternative of doing nothing."

In this article, we will walk through the actual science. We will explain what can go wrong when CRISPR edits human DNA, what clinical trial data show about real patients who have received gene editing therapies, how newer technologies are reducing risks, and what regulatory safeguards are in place to protect patients. The goal is to give you the information you need to form your own educated opinion, without either fear-mongering or false reassurance.

The Fundamental Risk: Off-Target Effects

When CRISPR-Cas9 edits DNA, it uses a short guide RNA to find a specific sequence in the genome. The Cas9 protein then cuts both strands of the DNA at that location. The cell's own repair machinery then fixes the break, either by stitching the ends together (a process called non-homologous end joining, or NHEJ) or by using a supplied DNA template to make a precise edit (homology-directed repair, or HDR).

The concern is that the guide RNA might direct Cas9 to the wrong place. The human genome contains about 3.2 billion base pairs, and many short sequences appear more than once. If a sequence elsewhere in the genome is similar enough to the intended target, Cas9 can bind there and cut. These unintended cuts are called off-target effects, and they are the most widely discussed safety concern in gene editing.

How Common Are Off-Target Effects?

Early CRISPR experiments, particularly those using first-generation guide RNAs and standard SpCas9, sometimes showed detectable off-target activity at dozens of genomic sites. This was alarming, and it fueled legitimate concern. However, the field has advanced considerably since those early experiments.

Modern guide RNA design uses computational tools that screen the entire genome for sequences similar to the intended target and select guides with the fewest potential off-target sites. High-fidelity Cas9 variants such as eSpCas9, HiFi Cas9, and SpCas9-HF1 have been engineered to be far more discriminating, cutting at the intended site while dramatically reducing activity at near-match sequences.

In practice, clinical-grade CRISPR therapies today are designed to have extremely low off-target activity. But "extremely low" is not zero, which is why rigorous detection methods matter.

How Off-Target Effects Are Detected

Researchers have developed several sophisticated methods to look for off-target cuts across the entire genome:

GUIDE-seq (Genome-wide Unbiased Identification of DSBs Evaluated by Sequencing): This method works by flooding cells with short double-stranded DNA tags. Wherever Cas9 creates a double-strand break, including at off-target sites, these tags integrate. The tagged sites are then sequenced, creating a genome-wide map of everywhere Cas9 actually cut. GUIDE-seq was one of the first unbiased methods and remains widely used.
CIRCLE-seq: This is an in vitro method that takes purified genomic DNA, circularizes it, and then exposes it to the Cas9-guide RNA complex. Any site that gets cut linearizes the circular DNA, and those fragments are sequenced. Because it uses purified DNA rather than living cells, CIRCLE-seq can detect even very low-frequency off-target sites. However, it tends to overestimate off-target risk because DNA in a test tube is more accessible than DNA wrapped in chromatin inside a living cell.
DISCOVER-seq (Discovery of In Situ Cas Off-targets and Verification by Sequencing): This newer method takes advantage of the cell's own DNA repair machinery. When Cas9 cuts DNA, repair proteins rush to the site. DISCOVER-seq uses ChIP-seq to identify where the MRE11 repair protein binds, effectively catching the cell in the act of repairing Cas9-induced breaks. It has the advantage of detecting off-target cuts as they actually occur inside living cells.
CHANGE-seq and rhAmpSeq: Additional methods that continue to improve sensitivity and reduce false positives.

Regulatory agencies such as the FDA require therapeutic developers to use multiple orthogonal methods to characterize off-target profiles before clinical trials proceed. This layered approach means that any meaningful off-target activity is highly likely to be detected during preclinical development.

Beyond Off-Targets: The Risks of Double-Strand Breaks

Off-target effects get the most attention, but there are other safety concerns inherent to any technology that deliberately cuts both strands of DNA. Even when CRISPR cuts at exactly the right place, the repair process itself can introduce problems.

Large Deletions

When Cas9 creates a double-strand break, the cell typically repairs it through NHEJ. This process is imprecise. While it usually produces small insertions or deletions (indels) of just a few base pairs at the cut site, studies have shown that it can occasionally cause much larger deletions spanning thousands or even millions of base pairs. A 2018 study from the Bhatt lab at the Wellcome Sanger Institute found that CRISPR editing in mouse embryonic stem cells sometimes produced deletions of several kilobases, and in some cases entire chromosome arms were affected.

These large deletions are concerning because they can disrupt neighboring genes that were not the intended target, potentially silencing tumor suppressors or activating oncogenes. Importantly, standard short-read sequencing methods may miss large deletions entirely. Long-read sequencing technologies such as PacBio and Oxford Nanopore are increasingly used to catch these events.

Chromosomal Rearrangements

When CRISPR makes cuts at two or more locations, whether on-target or off-target, the broken ends can rejoin incorrectly. This can produce chromosomal translocations, inversions, or other rearrangements. Chromosomal translocations are a hallmark of many cancers, so this is not a theoretical concern.

Multiplex editing strategies that cut at multiple sites simultaneously carry a higher risk of rearrangements than single-cut approaches. This is one reason that therapeutic applications tend to use a single guide RNA targeting one genomic site, minimizing the number of simultaneous breaks.

Chromothripsis

In rare cases, a single catastrophic event called chromothripsis can occur. Chromothripsis involves the shattering and random reassembly of a chromosome or chromosomal segment, producing dozens of rearrangements simultaneously. While chromothripsis has been observed in the context of CRISPR editing in laboratory studies, it appears to be a very low-frequency event. Nonetheless, it represents one of the more severe potential consequences of DNA double-strand breaks.

P53 Selection

One of the more subtle risks was highlighted in two high-profile 2018 papers published in Nature Medicine. Researchers found that cells with functional p53, the "guardian of the genome" tumor suppressor protein, tend to respond to CRISPR-induced double-strand breaks by activating cell death pathways. This means that in a population of edited cells, those with intact p53 are more likely to die, while cells with p53 mutations are more likely to survive and proliferate.

The practical implication is that CRISPR editing could theoretically enrich for cells with compromised p53, a protein whose loss is one of the most common drivers of cancer. This finding generated significant concern, although subsequent work has shown that the degree of selection depends heavily on the editing context, the cell type, and the magnitude of the DNA damage response. Clinical programs have since incorporated p53 pathway monitoring into their safety assessments, and no clinical evidence of p53-driven oncogenesis has emerged from CRISPR trials to date.

What Clinical Trials Actually Show

The most important safety data come not from laboratory experiments but from clinical trials in actual patients. As of early 2026, several CRISPR-based and gene editing therapies have generated substantial human safety data.

Casgevy (Exagamglogene Autotemcel) for Sickle Cell Disease and Beta-Thalassemia

Casgevy, developed by Vertex Pharmaceuticals and CRISPR Therapeutics, became the first CRISPR-based therapy to receive FDA approval in December 2023. It uses CRISPR-Cas9 to disrupt the BCL11A erythroid enhancer in a patient's own hematopoietic stem cells, reactivating fetal hemoglobin production.

The safety profile from the pivotal CLIMB SCD-121 and CLIMB Thal-111 trials has been encouraging. In sickle cell disease, 97% of evaluable patients achieved freedom from vaso-occlusive crises for at least 12 consecutive months. The most significant adverse events were related to the myeloablative conditioning regimen (high-dose busulfan chemotherapy) required to prepare the bone marrow for the edited cells, not to the gene editing itself. These conditioning-related side effects included neutropenia, thrombocytopenia, mucositis, febrile neutropenia, and hepatic veno-occlusive disease.

It is critical to distinguish between side effects caused by the chemotherapy conditioning and those attributable to the CRISPR editing. The conditioning regimen carries well-known risks, including infertility and secondary malignancies, that would be present with any bone marrow transplant procedure. As for the editing itself, extensive off-target analysis in Casgevy-treated cells showed no evidence of clinically significant off-target editing.

However, this does not mean long-term risks have been ruled out. The possibility that an undetected off-target event could contribute to cancer years or decades later cannot be completely excluded, which is why the FDA requires long-term follow-up (discussed below).

Intellia Therapeutics: NTLA-2001 for Transthyretin Amyloidosis

Intellia's NTLA-2001 represents a different paradigm: it is the first CRISPR therapy delivered in vivo, meaning the editing components are injected directly into the patient's bloodstream rather than applied to cells in a laboratory. Encapsulated in lipid nanoparticles that preferentially accumulate in the liver, NTLA-2001 uses CRISPR to knock out the transthyretin (TTR) gene in hepatocytes, reducing production of the misfolded protein that causes transthyretin amyloidosis (ATTR).

In clinical trials, NTLA-2001 has demonstrated significant TTR protein reduction, with some patients achieving greater than 90% knockdown that has been sustained over multi-year follow-up. The safety signals have been manageable. The most commonly reported adverse events have included:

Transient infusion-related reactions: These have been mild to moderate in most patients, consistent with the immune system's response to lipid nanoparticles. Symptoms include flushing, increased heart rate, and mild changes in blood pressure.
Liver enzyme elevations: Transient increases in alanine aminotransferase (ALT) and aspartate aminotransferase (AST) have been observed, which is expected given that the therapy directly targets liver cells. These elevations have generally resolved without clinical consequences.

The in vivo delivery approach introduces its own set of considerations compared to ex vivo editing. Once the editing components are in the body, there is less control over which cells get edited and how long the Cas9 protein remains active. However, lipid nanoparticles are cleared relatively quickly, limiting the duration of Cas9 exposure and thereby reducing the window for off-target activity.

Verve Therapeutics: VERVE-101 for Familial Hypercholesterolemia

Verve's VERVE-101 program targets the PCSK9 gene in the liver using a base editing approach to permanently lower LDL cholesterol in patients with heterozygous familial hypercholesterolemia. While base editing (discussed in more detail below) avoids double-strand breaks, the VERVE-101 program has encountered meaningful safety signals that deserve honest discussion.

In the heart-1 clinical trial, some patients experienced:

Grade 3 ALT elevations: More significant liver enzyme increases than seen with NTLA-2001, indicating a greater degree of liver stress in some patients.
Thrombocytopenia: Decreases in platelet counts, which can increase bleeding risk.
One patient death: A patient in the trial died from a cardiac event. Verve and the Data Safety Monitoring Board assessed this as related to the patient's underlying severe cardiovascular disease rather than the therapy, but the event underscored the complexity of treating critically ill patients with novel therapies. The death was a sobering reminder that clinical context matters and that early-stage trials in patients with advanced disease carry inherent risks.

Verve has since refined its dosing strategy and patient selection criteria. The program continues, but the experience illustrates that even newer editing approaches are not without risk, particularly when targeting organs under significant disease burden.

Gene Therapy Delivery: The AAV Safety Question

While not specific to CRISPR, the safety of gene therapy delivery vehicles is inseparable from the overall safety equation. Many gene therapies use adeno-associated virus (AAV) vectors to deliver genetic material to target cells. AAV-based therapies have their own distinct risk profile that is important to understand.

Hepatotoxicity

AAV vectors, particularly those that target the liver, can trigger severe immune responses against transduced hepatocytes. This hepatotoxicity has been a recurring concern across multiple AAV gene therapy programs. At high doses, the immune response can cause significant liver damage, and managing this requires careful dose selection and immunosuppressive regimens.

The Audentes/Astellas Deaths

Perhaps the most serious safety event in the gene therapy field involved Audentes Therapeutics' AT132 program for X-linked myotubular myopathy. Several patients, all young boys with the severe form of the disease, died after receiving high doses of AAV8 vector. The deaths were attributed to a combination of very high vector doses and pre-existing liver disease in the affected patients. Subsequent investigation pointed to hepatotoxicity and a sepsis-like inflammatory syndrome.

These deaths led to a fundamental reassessment of AAV dosing strategies across the gene therapy field. They also highlighted the importance of understanding how pre-existing conditions affect the risk profile of gene therapy.

Thrombotic Microangiopathy (TMA)

Another serious adverse event seen with high-dose AAV therapy is thrombotic microangiopathy (TMA), a condition involving small blood vessel damage, low platelet counts, and organ injury. TMA has been reported in several AAV clinical programs and appears to be dose-dependent. The mechanism likely involves complement activation triggered by the large viral load.

These AAV safety concerns are relevant to the CRISPR discussion because the delivery vehicle is often as important as the editing payload. The shift toward lipid nanoparticle delivery (as used by Intellia and Verve) partly reflects an effort to avoid AAV-associated risks, though LNPs introduce their own set of considerations.

How Newer Editing Technologies Reduce Risk

Recognizing that double-strand breaks are the root cause of many safety concerns, researchers have developed editing technologies that modify DNA without cutting both strands.

Base Editing: Chemical Conversion Without Cutting

Base editors, pioneered by David Liu's lab at the Broad Institute, use a modified Cas9 that is catalytically impaired (it cannot cut both DNA strands) fused to a deaminase enzyme. The deaminase chemically converts one DNA base to another at the target site. Cytosine base editors (CBEs) convert C-G base pairs to T-A, while adenine base editors (ABEs) convert A-T to G-C.

Because base editors do not create double-strand breaks, they largely eliminate the risks of large deletions, chromosomal rearrangements, chromothripsis, and p53 selection. This is a significant safety advantage and is one reason why base editing has attracted enormous interest for therapeutic applications.

However, base editing introduces its own set of off-target concerns:

DNA off-targets: Like CRISPR-Cas9, the guide RNA can direct the base editor to the wrong genomic location. At those off-target sites, the deaminase can convert bases, creating unwanted point mutations rather than double-strand breaks. These are generally considered less damaging than DSBs, but a point mutation in a critical gene could still be harmful.
Bystander editing: If there are multiple target bases within the editing window (typically positions 4-8 of the protospacer), the deaminase may edit unintended nearby bases in addition to the desired one.
Guide-independent DNA off-targets: Some early base editors showed deaminase activity at sites not directed by the guide RNA, essentially random C-to-U or A-to-I conversions across the genome. Engineered variants have substantially reduced this activity.

RNA Off-Targets from Base Editing

A particularly important concern for adenine base editors is transcriptome-wide RNA off-target activity. The adenosine deaminase used in ABEs, particularly the highly active ABE8e variant, can deaminate adenosine to inosine not only in DNA but also in RNA transcripts throughout the cell.

A 2019 study by Grunewald et al. in Nature demonstrated that both CBEs and ABEs could cause widespread off-target editing in the transcriptome, with ABEs converting thousands of adenosines to inosines across cellular RNA. While RNA is transient and these changes would not be heritable, widespread RNA editing could transiently alter protein function across the cell in unpredictable ways.

Subsequent engineering efforts have produced ABE variants with reduced RNA editing activity, such as ABE8e-V106W, which maintains high on-target DNA editing efficiency while substantially lowering transcriptome-wide RNA deamination. This remains an active area of optimization.

Prime Editing: Search and Replace

Prime editing, also developed by David Liu's group, uses a Cas9 nickase (which cuts only one DNA strand) fused to a reverse transcriptase. A prime editing guide RNA (pegRNA) both specifies the target site and encodes the desired edit. The reverse transcriptase copies the edit from the pegRNA directly into the target strand.

Because prime editing nicks only one strand rather than creating a double-strand break, it avoids the most serious DSB-associated risks. It can make all 12 types of point mutations as well as small insertions and deletions, giving it greater versatility than base editing. Off-target activity with prime editing has generally been very low in published studies.

Prime editing is less efficient than CRISPR-Cas9 nucleases, particularly for edits in certain genomic contexts, but its safety profile makes it attractive for therapeutic applications where precision matters more than editing rates.

Regulatory Safeguards: How the System Protects Patients

The regulatory framework surrounding gene editing therapies involves multiple layers of oversight designed to catch safety problems before, during, and after clinical use.

Preclinical Requirements

Before any CRISPR therapy enters human trials, developers must submit an Investigational New Drug (IND) application to the FDA that includes:

Comprehensive off-target analysis using multiple detection methods
Genotoxicity studies assessing chromosomal integrity
Biodistribution studies showing where the editing components go in the body
Animal toxicology studies, often in multiple species
Manufacturing data demonstrating consistency and purity of the editing components

Clinical Trial Oversight

During clinical trials, independent Data Safety Monitoring Boards (DSMBs) review safety data on an ongoing basis. DSMBs have the authority to pause or halt trials if concerning safety signals emerge. They operate independently from both the sponsor and the FDA, providing an additional check against conflicts of interest.

Dose escalation in early trials follows a cautious approach, starting with the lowest expected therapeutic dose and increasing only after safety review of each cohort.

The 15-Year Follow-Up Requirement

For gene therapies, including CRISPR-based treatments, the FDA requires manufacturers to conduct long-term follow-up studies lasting at least 15 years after treatment. This requirement reflects the reality that some risks, particularly cancer risk from insertional mutagenesis or off-target effects, may not become apparent for years or even decades.

During this follow-up period, patients undergo regular monitoring for:

Malignancies, particularly hematologic cancers for therapies that edit blood stem cells
Autoimmune disorders
New or worsening neurological conditions
Unexpected changes in blood counts or organ function

This 15-year monitoring window is a critical safety net. It acknowledges the limits of what preclinical testing and short-term clinical trials can reveal about technologies that permanently alter a patient's genome.

Post-Market Surveillance

Once a gene therapy is approved, the FDA can require Risk Evaluation and Mitigation Strategies (REMS) that mandate specific monitoring, prescriber training, or patient registries. The agency also monitors adverse event reports through the MedWatch system and can take regulatory action if new safety signals emerge.

Germline vs. Somatic Editing: A Bright Line

All approved and in-development CRISPR therapies target somatic cells, meaning the edits affect only the treated patient and are not passed on to future generations. This is a critical distinction.

Somatic editing changes DNA in specific tissues or cell types in an individual. If the patient's liver cells are edited, those changes exist only in the patient's liver and cannot be inherited by their children.

Germline editing would alter DNA in embryos, eggs, or sperm, meaning the changes would be passed to all future generations. Germline editing in humans is effectively banned worldwide. In the United States, Congress has prohibited the FDA from even reviewing applications for clinical trials involving germline editing. Most other countries have similar prohibitions or moratoriums.

The reason for this bright line is straightforward: germline edits are irreversible at a population level. If an edit turns out to have unforeseen consequences, those consequences would propagate through future generations. The scientific community broadly agrees that the technology is not ready for germline applications and that the ethical questions surrounding deliberate modification of the human gene pool require far more societal deliberation.

The case of He Jiankui, the Chinese researcher who created the first gene-edited babies in 2018 using CRISPR to modify the CCR5 gene, serves as a cautionary example. His work was widely condemned by the scientific community as reckless, premature, and ethically unacceptable. He was sentenced to three years in prison by Chinese authorities. The babies, now children, will require lifelong monitoring for potential health consequences from edits whose full effects remain unknown.

Putting It All Together: The Honest Risk Assessment

So, is CRISPR safe? Here is what an honest, evidence-based assessment looks like as of early 2026:

What the data support

Casgevy has shown a strong clinical safety profile across multiple years of follow-up, with the majority of adverse events attributable to the required chemotherapy conditioning rather than the gene editing itself. The benefit for patients with severe sickle cell disease, who face a life of pain crises and organ damage, clearly outweighs the known risks.
In vivo CRISPR delivery via LNPs (Intellia's NTLA-2001) has been well tolerated in clinical trials, with infusion reactions and transient liver enzyme elevations as the primary adverse events. These have been manageable and reversible.
Off-target editing in clinical-grade therapies has been extremely low as measured by current detection methods. No clinical trial has reported evidence of an off-target event causing patient harm.
No CRISPR-related cancers have been reported in any clinical trial to date, despite the theoretical concerns about large deletions, rearrangements, and p53 selection.

What remains uncertain

Long-term cancer risk cannot yet be fully assessed. The 15-year follow-up requirement exists precisely because some risks may take years to manifest. We do not yet have decade-long data on CRISPR-treated patients.
In vivo delivery to organs beyond the liver remains challenging, and safety profiles may differ for different target tissues and delivery methods.
The Verve VERVE-101 experience shows that safety concerns are real, particularly when treating patients with severe underlying disease. The therapeutic window for some applications may be narrower than initially hoped.
Detection methods, while vastly improved, may still miss rare off-target events. An off-target edit occurring in 1 in 10,000 cells might not be detected by current sequencing approaches but could still be clinically significant if it occurs in a stem cell with proliferative advantage.

What is clearly improving

High-fidelity Cas9 variants have dramatically reduced off-target cutting compared to first-generation tools.
Base editing and prime editing eliminate double-strand break risks entirely, addressing the most serious mechanistic safety concerns.
Delivery technologies are becoming more targeted, reducing exposure of non-target tissues.
Detection methods continue to improve in sensitivity and accuracy, meaning safety profiles are characterized more thoroughly than ever before.
Regulatory frameworks have matured significantly, with clear guidelines for preclinical testing, clinical monitoring, and long-term follow-up.

The Risk-Benefit Equation

Safety in medicine is never absolute. Every drug, every surgery, every medical intervention carries risk. The relevant question is whether the benefits outweigh the risks for a given patient population.

For a patient with severe sickle cell disease who experiences frequent pain crises, progressive organ damage, and a shortened life expectancy, the risk-benefit calculation for Casgevy is compelling. The known risks of the procedure, primarily from conditioning chemotherapy, are real but manageable, and the potential benefit is transformative.

For a patient with mild hypercholesterolemia who could be adequately treated with a statin, the calculus would be very different. The risks of a novel in vivo gene editing therapy would be harder to justify when safer alternatives exist.

This is how medicine has always worked. The first heart transplant carried enormous risk, but for patients with terminal heart failure, it represented their only chance. As the technology matured, risks decreased and the range of eligible patients expanded. Gene editing is on a similar trajectory.

Looking Forward

The safety profile of CRISPR and gene editing technologies is not a fixed quantity. It is improving with each generation of tools, each clinical trial, and each advance in our ability to detect and prevent adverse events. The researchers and clinicians working in this field are acutely aware of the risks and are working systematically to address them.

What patients and the public deserve is honest communication about both the promise and the limitations of these technologies. Gene editing is not risk-free, but for serious genetic diseases with limited treatment options, it offers something that was not possible a decade ago: the prospect of a durable correction at the level of DNA itself.

The next few years will be decisive. As more patients are treated and longer-term follow-up data accumulate, the safety picture will come into sharper focus. Until then, the appropriate posture is one of cautious optimism, grounded in data rather than hype.

Sources and Further Reading

Tsai, S.Q. et al. "GUIDE-seq enables genome-wide profiling of off-target cleavage by CRISPR-Cas nucleases." Nature Biotechnology 33, 187-197 (2015).
Tsai, S.Q. et al. "CIRCLE-seq: a highly sensitive in vitro screen for genome-wide CRISPR-Cas9 nuclease off-targets." Nature Methods 14, 607-614 (2017).
Wienert, B. et al. "Unbiased detection of CRISPR off-targets in vivo using DISCOVER-Seq." Science 364, 286-289 (2019).
Kosicki, M., Tomberg, K. & Bradley, A. "Repair of double-strand breaks induced by CRISPR-Cas9 leads to large deletions and complex rearrangements." Nature Biotechnology 36, 765-771 (2018).
Haapaniemi, E. et al. "CRISPR-Cas9 genome editing induces a p53-mediated DNA damage response." Nature Medicine 24, 927-930 (2018).
Ihry, R.J. et al. "p53 inhibits CRISPR-Cas9 engineering in human pluripotent stem cells." Nature Medicine 24, 939-946 (2018).
Grunewald, J. et al. "Transcriptome-wide off-target RNA editing induced by CRISPR-guided DNA base editors." Nature 569, 433-437 (2019).
Frangoul, H. et al. "CRISPR-Cas9 Gene Editing for Sickle Cell Disease and Beta-Thalassemia." New England Journal of Medicine 384, 252-260 (2021).
Gillmore, J.D. et al. "CRISPR-Cas9 In Vivo Gene Editing for Transthyretin Amyloidosis." New England Journal of Medicine 385, 493-502 (2021).
Rees, H.A. & Liu, D.R. "Base editing: precision chemistry on the genome and transcriptome of living cells." Nature Reviews Genetics 19, 770-788 (2018).
Anzalone, A.V. et al. "Search-and-replace genome editing without double-strand breaks or donor DNA." Nature 576, 149-157 (2019).
FDA Guidance: "Long Term Follow-Up After Administration of a Gene Therapy Product." U.S. Food and Drug Administration (2020).
Vertex Pharmaceuticals. Casgevy prescribing information and clinical trial data (2023-2025).
Intellia Therapeutics. NTLA-2001 clinical trial updates, ASH and AAN presentations (2021-2025).
Verve Therapeutics. heart-1 trial data presentations and safety updates (2023-2025).