Literature horizon scan for new scientific data on plants, microorganisms and animals, and their products obtained by new genomic techniques (March 2026) Authors: European Food Safety Authority (EFSA), Michele Ardizzone, Fulvio Barizzone, Martina Bonatti, Alice Branchi, Tilemachos Goumperis, Dafni Maria Kagkli, Paolo Lenzi, Aleksandra Lewandowska, Ana M. Camargo, Irene Pilar Munoz Guajardo, Nikoletta Papadopoulou, Tommaso Raffaello Published: 2026 ABSTRACT AbstractThe European Food Safety Authority has issued several scientific opinions on plants, microorganisms and animals obtained through certain new genomic techniques (NGTs), following requests received by the European Commission. These scientific opinions provided considerations on the potential risks associated with NGTs, as compared to conventional breeding techniques and established genomic techniques (EGTs), and on the applicability of existing guidelines for the risk assessment of plants, microorganisms and animals obtained by NGTs and products thereof. Against this background, EFSA was asked by the European Commission ‘to provide scientific and technical assistance for a regular horizon scanning to assess new scientific data on plants, microorganisms and animals, and products thereof obtained by new genomic techniques’, to assess any new data and evidence emerging from these studies, and to consider whether it may have implications for EFSA's relevant scientific opinions. A critical assessment of the quality and relevance of these studies should also be conducted, and biannual reports delivered to the Commission. This report presents the outcome of the literature search on new scientific data on plants, microorganisms and animals and products thereof obtained by NGTs, performed as described in the protocol made available for public consultation in May–June 2025. The report describes search strategies and inclusion/exclusion criteria applied for literature review and presents the results of the assessment performed for new scientific data (published up to March 2026) on plants, microorganisms and animals and products thereof obtained by NGTs. The report discusses the limitations of the search and provides recommendations for improvement. EFSA concluded that none of the studies retrieved by the literature search contained new hazards or risks not previously considered in EFSA scientific opinions. INTRODUCTION Background and Terms of Reference (ToR) as provided by the requestor 1 Background Given the rapid development of new genomic techniques (NGTs) and their potential applications, new scientific studies are regularly published regarding the development and current or potential market presence of NGTs in plants, animals and microorganisms. These studies can contain new scientific data and evidence relevant to the safety, risk assessment and other considerations made by EFSA in its adopted or upcoming scientific opinions on NGTs applied to plants, microorganisms and animals. Beyond the extensive literature reviews conducted by EFSA in response to various mandates in this domain, EFSA has also independently analysed new evidence related to its scientific work on organisms obtained by NGTs. In addition, the European Commission (EC) has requested EFSA to consider individual scientific publications brought to its attention by stakeholders on an ‘ad‐hoc’ basis. For reasons of completeness and transparency, a more formal and structured approach, including biannual reporting to the EC, is deemed necessary. Terms of Reference as provided by the requestor In light of the above, in accordance with Article 31 of Regulation (EC) No 178/2002, EFSA is requested to regularly screen the scientific literature to identify relevant studies for the assessment of food and feed safety and the environmental safety of organisms and products obtained with NGTs, to assess any new evidence emerging from these studies, and to consider whether it may have implications for EFSA's relevant scientific opinions. A critical assessment of the quality and relevance of these studies should also be conducted, and regular reports delivered to the Commission. This request does not include evidence assessed in adopted and upcoming opinions. More specifically, EFSA is requested to: Develop and validate search strategies to be used for searching of scientific literature databases with adequate sensitivity and specificity to identify studies that could be relevant as regards the safety, risk assessment and other considerations examined by EFSA in its relevant scientific opinions concerning NGTs applied to plants, micro‐organisms and animals.Develop and pilot criteria for evidence inclusion and exclusion, and critical appraisal tools to support the assessment of relevant published scientific evidence.Extract and summarise relevant data and evidence and assess whether any new data and evidence may have implications for EFSA relevant scientific opinions.Deliver biannual reports on relevant findings. Develop and validate search strategies to be used for searching of scientific literature databases with adequate sensitivity and specificity to identify studies that could be relevant as regards the safety, risk assessment and other considerations examined by EFSA in its relevant scientific opinions concerning NGTs applied to plants, micro‐organisms and animals. Develop and validate search strategies to be used for searching of scientific literature databases with adequate sensitivity and specificity to identify studies that could be relevant as regards the safety, risk assessment and other considerations examined by EFSA in its relevant scientific opinions concerning NGTs applied to plants, micro‐organisms and animals. Develop and pilot criteria for evidence inclusion and exclusion, and critical appraisal tools to support the assessment of relevant published scientific evidence. Develop and pilot criteria for evidence inclusion and exclusion, and critical appraisal tools to support the assessment of relevant published scientific evidence. Extract and summarise relevant data and evidence and assess whether any new data and evidence may have implications for EFSA relevant scientific opinions. Extract and summarise relevant data and evidence and assess whether any new data and evidence may have implications for EFSA relevant scientific opinions. Deliver biannual reports on relevant findings. Deliver biannual reports on relevant findings. Interpretation of the Terms of Reference EFSA interprets the overall objective of the mandate as follows:To identify and assess new evidence relevant for the safety, risk assessment and other considerations examined by EFSA in its scientific opinions on NGTs applied to plants, microorganisms and animals. To identify and assess new evidence relevant for the safety, risk assessment and other considerations examined by EFSA in its scientific opinions on NGTs applied to plants, microorganisms and animals. This objective was translated into assessment questions (AQs) and sub‐questions (SQs) as detailed in the literature search protocol (Annex A). The protocol describes the search strategies, inclusion and exclusion criteria, methodology for evidence appraisal and lists the ‘EFSA's relevant scientific opinions’ defined by the mandate, together with their conclusions (Appendix A to the protocol). The draft protocol was shared for public consultation. To address the mandate, the following NGT definition was used: techniques capable to change the genetic material of an organism and that have emerged or have been developed since 2001, when the existing GMO legislation 2 was adopted. This definition was used by the European Commission in ‘Study on the status of new genomic techniques under Union law and in light of the Court of Justice ruling in Case C‐528/16 3 ’. This report is the third biannual report requested in ToR4, describes the modifications in evidence appraisal process (ToR2) and presents the results for all AQs and their sub‐questions. AQ1: Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to plants? SQ1.1: Is there evidence of new hazards or risks not previously considered in EFSA scientific opinions? SQ1.2: Does the new evidence have any implications for EFSA's relevant scientific opinions on NGTs applied to plants? AQ2: Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to microorganisms? SQ2.1: Is there evidence for new hazards or risks not previously considered in EFSA scientific opinions? SQ2.2: Does the new evidence have any implications for EFSA's relevant scientific opinions on NGTs applied to microorganisms? AQ3: Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to animals? SQ3.1: Is there evidence for new hazards or risks not previously considered in EFSA scientific opinions? SQ3.2: Does the new evidence have any implications for EFSA's relevant scientific opinions on NGTs applied to animals? DATA AND METHODOLOGIES Data EFSA was requested to consider any new evidence emerging from the literature search with potential impact on the risk assessment considerations included in past EFSA scientific opinions on plants, microorganisms and animals obtained by NGTs. Table 1 reports the opinions that have been published by EFSA in this field. TABLE 1EFSA scientific opinions relevant for each AQ.ReferenceTitle AQ1 (NGTs applied to plants) EFSA GMO Panel (2012a)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms). (2012a). Scientific opinion addressing the safety assessment of plants developed using Zinc Finger Nuclease 3 and other Site‐Directed Nucleases with similar function. EFSA Journal, 10(10), 2943. https://doi.org/10.2903/j.efsa.2012.2943 EFSA GMO Panel (2012b)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms). (2012b). Scientific opinion addressing the safety assessment of plants developed through cisgenesis and intragenesis. EFSA Journal, 10(2), 2561. https://doi.org/10.2903/j.efsa.2012.2561 EFSA GMO Panel (2020)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms). (2020). Applicability of the EFSA Opinion on SDNs type 3 for the safety assessment of plants developed using SDNs type 1 and 2 and oligonucleotide‐directed mutagenesis. EFSA Journal, 18(11), 6299. https://doi.org/10.2903/j.efsa.2020.6299 EFSA GMO Panel (2021a)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms). (2021a). Scientific Opinion on the evaluation of existing guidelines for their adequacy for the molecular characterisation and environmental risk assessment of genetically modified plants obtained through synthetic biology. EFSA Journal, 19(2), 6301. https://doi.org/10.2903/j.efsa.2021.6301 EFSA GMO Panel (2022a)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms). (2022a). Updated scientific opinion on plants developed through cisgenesis and intragenesis. EFSA Journal, 20(10), 7621. https://doi.org/10.2903/j.efsa.2022.7621 EFSA GMO Panel (2022b)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms). (2022b). Scientific Opinion on the evaluation of existing guidelines for their adequacy for the food and feed risk assessment of genetically modified plants obtained through synthetic biology. EFSA Journal, 20(7), 7410. https://doi.org/10.2903/j.efsa.2022.7410 EFSA GMO Panel (2022c)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms). (2022c). Statement on criteria for risk assessment of plants produced by targeted mutagenesis, cisgenesis and intragenesis. EFSA Journal 2022;20(10):7618, 12 pp. https://doi.org/10.2903/j.efsa.2022.7618 AQ2 (NGTs applied to microorganisms) EFSA Scientific Committee (2020)EFSA Scientific Committee. (2020). Scientific Opinion on the evaluation of existing guidelines for their adequacy for the microbial characterisation and environmental risk assessment of microorganisms obtained through synthetic biology. EFSA Journal, 18(10), 6263. https://doi.org/10.2903/j.efsa.2020.6263 EFSA Scientific Committee (2022)EFSA Scientific Committee. (2022). Scientific Opinion on the evaluation of existing guidelines for their adequacy for the food and feed risk assessment of microorganisms obtained through synthetic biology. EFSA Journal, 20(8), 7479. https://doi.org/10.2903/j.efsa.2022.7479 EFSA GMO Panel (2024)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms), Mullins, E., Bresson, J.‐L., Dewhurst, I. C., Epstein, M. M., Firbank, L. G., Guerche, P., Hejatko, J., Moreno, F. J., Naegeli, H., Nogué, F., Rostoks, N., Sánchez Serrano, J. J., Savoini, G., Veromann, E., Veronesi, F., Cocconcelli, P. S., Glandorf, D., Herman, L., … Dalmay, T. (2024). New developments in biotechnology applied to microorganisms. EFSA Journal, 22(7), e8895. https://doi.org/10.2903/j.efsa.2024.8895 AQ3 (NGTs applied to animals) EFSA GMO Panel (2025)EFSA GMO Panel (EFSA Panel on Genetically Modified Organisms), Casacuberta, J., Barro, F., Braeuning, A., de Maagd, R., Epstein, M. M., Frenzel, T., Gallois, J.‐L., Koning, F., Messéan, A., Moreno, F. J., Nogué, F., Schulman, A. H., Tebbe, C., Veromann, E., Firbank, L., Glandorf, D., Herskin, M. S., Lillico, S. G., … Savoini, G. (2025). New developments in biotechnology applied to animals: An assessment of the adequacy and sufficiency of current EFSA guidance for animal risk assessment. EFSA Journal, 23(8), e9566. https://doi.org/10.2903/j.efsa.2025.9566 EFSA scientific opinions relevant for each AQ. Methodologies Problem formulation A detailed description of the problem formulation is reported in the protocol that is included in Annex A. The problem formulation has been made operational by translating the ToRs into assessment questions, in line with the ‘Guidance on protocol development for EFSA generic scientific assessments’ (EFSA Scientific Committee, 2023). Literature search EFSA developed search strings and inclusion/exclusion criteria assisted by a multidisciplinary group of scientists, and ran the searches followed by analysis of the evidence. The experts from the GMO Panel's working groups (WGs) on Molecular Characterisation (MC WG) and the Cross‐Cutting WG (XC WG) were consulted on the search strings, as well as on the scientific publications identified for plants, microorganisms and animals obtained by NGTs, as potentially relevant to the mandate ToRs. More details on the search methodology are included in the protocol in Annex A. All search strings for AQ1‐3, presented in tables 5–13 of the protocol, are constructed from four key terms: New genomic techniques AND Safety AND Species AND Time limits. The NGT terms used cover a variety of expressions that are interchangeably used when mentioning NGTs as defined in Section 1.2, including terms related to specific techniques such as prime editing, base editing, TALEN, ZFN etc. For the term Species, the terms were based on those used in EFSA GMO Panel (2022a) for plants, EFSA GMO Panel (2024) for microorganisms and Van Eenennaam (2023) for animals, taking into the account additional terms suggested by public consultation. The time limits for each of the searches and the dates of searches are presented in Table 2. TABLE 2Time ranges and dates of literature searches performed using three different sources of evidence.AQTime range for the searchDate of searchAQ101/01/2022 – onwards11/3/2026AQ201/01/2024 – onwardsAQ301/01/2024 – onwards Time ranges and dates of literature searches performed using three different sources of evidence. Consultation In line with its policy on openness and transparency, EFSA consulted EU Member States and its stakeholders by an online public consultation. All stakeholders were invited to submit their comments on the draft protocol between May 2025 and June 2025. 4 Following this consultation process, the document was revised by EFSA and consulted with the XC WG. The outcome of the online public consultation is reported in Annex which is published on EFSA's website together with the first biannual report (EFSA, 2025) and was taken into account when amending the protocol (Annex A). ASSESSMENT The number of publications retrieved after each step of the literature search is presented on Figures 1, 2, 3, for AQs 1–3, respectively. The outcome for each AQ is presented as a PRISMA diagram (Page et al., 2021). These results are discussed in Section 3.1, ToR1, the criteria for inclusion and exclusion in Section 3.2, ToR2 and the data and evidence in retrieved publications in Section 3.3, ToR3. Number of publications retrieved after each step of the literature search for AQ1. Number of publications retrieved after each step of the literature search for AQ2. Number of publications retrieved after each step of the literature search for AQ3. ToR1. Develop and validate search strategies to be used for searching of scientific literature databases with adequate sensitivity and specificity to identify studies that could be relevant as regards the safety, risk assessment and other considerations examined by EFSA in its relevant scientific opinions concerning NGTs applied to plants, microorganisms and animals The search strategies and reference publications for all AQs used for identifying search terms as well as validating the search strategy are the same as those described in the previous report for this mandate (EFSA, 2026). Minor inconsistencies in the search terms were amended and listed as changes in the protocol (Annex A), and are not expected to affect the outcome of the searches. The searches covered the entire time range of the horizon scan per organism, followed by the removal of all previously assessed references, to exclude the possibility of missing any references affected by the modification of the search terms since then. The search terms could be further updated to include new terms when needed as the horizon scanning methodology evolves through repeated testing, and as the scientific knowledge progresses. For all three AQs, it was observed that some publications were identified by search strings designed for one of the other groups of organisms, e.g. a publication potentially relevant for microorganisms was retrieved only by a search string for animals. Such publications were transferred to the more appropriate reference pool and screened again by two reviewers for their relevance. These number of publications transferred between respective groups is reported on Figures 1, 2, 3. Overall, the search strategies for all 3 AQs are of satisfactory sensitivity and specificity since (i) all reference publications were identified and (ii) low number of retrieved publications with no relevance to the topic of NGTs were retrieved. For all AQs, the possible limitations of the search strategies, related to the key terms described in Section 2.2.2, were described in the 1st report of this mandate (EFSA, 2025) and are still applicable to this report. AQ1. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to plants? Except for minor editorial changes not affecting the sensitivity of the search, the search strategy for AQ1 was not modified in comparison to that of the previous report (EFSA, 2026). AQ2. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to microorganisms? Except for minor editorial changes not affecting the sensitivity of the search, the search strategy for AQ1 was not modified in comparison to that of the previous report (EFSA, 2026). As described in the previous report, it was tested whether the use of the word ‘human’ in the title or abstract of a publication suggesting medical applications that are out of scope of this mandate, can be linked to the relevance of retrieved publications. It was concluded that this term will not be used to exclude publications a priori, not to compromise the sensitivity of the search. Consequently, the search terms no longer distinguish between the records containing the term ‘human’ in their title or abstract and those that do not. AQ3. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to animals? Except for minor editorial changes not affecting the sensitivity of the search, the search strategy for AQ1 was not modified in comparison to that of the previous report (EFSA, 2026). As described in the previous report, it was tested whether the relevance of the retrieved studies for AQ3 correlates to the species mentioned in the title/abstract (either agri‐food or model animals). It was hypothesised that studies focusing on model animals might be more often linked to early‐stage studies (including studies on mammalian cell lines) and/or medical applications and therefore less relevant for the mandate. It was concluded that studies performed on both groups might contain relevant information and the model animals will not be excluded from the search terms in the upcoming reports for this mandate. Consequently, the search terms no longer distinguish studies from both animal groups. ToR2. Develop and pilot criteria for evidence inclusion and exclusion, and critical appraisal tools to support the assessment of relevant published scientific evidence Possible limitations of the used inclusion/exclusion criteria were described in the previous reports for this mandate (EFSA, 2025, 2026), and all previous considerations apply to the criteria used for all AQs (see Annex A, Tables 2 and 3). The artificial intelligence (AI) functionality within DistillerSR (DistillerSR AI Screening function) (Cagnoni et al., 2023) was used to support the title and abstract screening phase by acting as a second reviewer. The algorithm was trained on titles and abstracts screened in the first two horizon‐scanning reports (EFSA, 2025, 2026) prepared under this mandate, for a total of 4472 references reviewed in duplicate (122 included and 4350 excluded). 15 repetitions made by a random sample of 80% of the included references and by a random sample of excluded references of a size of 5 times the sample of included references were taken. Each time the algorithm was trained on the references identified and validated against the remaining ones. The overall sensitivity of the model was above 0.92 and the specificity about 0.94. The algorithm was considered as appropriate for the task and it was applied separately to each search string (AQ1, AQ2, AQ3). Each record was therefore assessed by one human reviewer and the DistillerSR AI Screening function; where disagreements occurred, these were evaluated by an additional EFSA reviewer and resolved through discussion. This process was consistently applied across all search strings. AQ1. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to plants? The performance of AI as a second reviewer had a low disagreement rate between human and AI assessments (43/650 records, 6.6%). When discrepancies occurred, they were evaluated by an additional EFSA reviewer and in 28/43 (65%) of the cases the human assessment was confirmed. Among the conflicting cases, 12 (12/43, 27.9%) were identified as false negatives, although only one of these (Juhas et al., 2025) was considered relevant following full‐text screening. The conflicts by exclusion criterion are summarised in Table 3. Most disagreements were associated with the criterion ‘Outcome/conclusions’, while fewer conflicts concerned the organism, publication type or technique. The low number of false negatives deemed relevant after full‐text screening indicates that the use of AI as a second reviewer supports the identification of relevant evidence for AQ1. TABLE 3Conflict resolution between human and AI reviewer for AQ1.Criteria for exclusionFalse positives a False negatives b Agreement with AI reviewerAgreement with human reviewerAgreement with AI reviewerAgreement with human reviewerOrganism61Outcome/conclusions36911Publication type61Technique11Total: 19Total: 24 Note: Each record could have had more than one criteria for exclusion. a ‘False positives’ are the records which the human reviewer scored as not relevant, while AI scored it as relevant. b ‘False negatives’ are the records which the human reviewer scored as relevant, while AI scored it as not relevant. Conflict resolution between human and AI reviewer for AQ1. Note: Each record could have had more than one criteria for exclusion. ‘False positives’ are the records which the human reviewer scored as not relevant, while AI scored it as relevant. ‘False negatives’ are the records which the human reviewer scored as relevant, while AI scored it as not relevant. The criteria for evidence inclusion and exclusion used for AQ1 are the same as reported in the first and second report for this mandate (EFSA, 2025, 2026) and all previous considerations apply. AQ2. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to microorganisms? The performance of AI as a second reviewer had a low disagreement rate between human and AI assessments (20/597 records, 3.4%). Discrepancies were evaluated by an additional EFSA reviewer, and in 16/20 (80.0%) of the cases the human assessment was confirmed. Among the conflicting cases, 2 (2/20, 10%) were identified as false negatives and none of these was later deemed relevant following full‐text screening. As shown in Table 4, most conflicts were associated with the criterion ‘Outcome/conclusions’, whereas a smaller number related to organism, scope/intended use or technique. The outcome indicates that AI performed well as a second reviewer for AQ2 and that the few false negatives identified did not affect the final set of relevant publications. TABLE 4Conflict resolution between human and AI reviewer for AQ2.Criteria for exclusionFalse positives a False negatives b Agreement with AI reviewerAgreement with AI reviewerAgreement with human reviewerAgreement with AI reviewerAgreement with human reviewerOrganism2Outcome/conclusions1922Scope/intended use3Technique1Total: 15Total: 5 Note: Each record could have had more than one criteria for exclusion. a ‘False positives’ are the records which the human reviewer scored as not relevant, while AI scored it as relevant. b ‘False negatives’ are the records which the human reviewer scored as relevant, while AI scored it as not relevant. Conflict resolution between human and AI reviewer for AQ2. Note: Each record could have had more than one criteria for exclusion. ‘False positives’ are the records which the human reviewer scored as not relevant, while AI scored it as relevant. ‘False negatives’ are the records which the human reviewer scored as relevant, while AI scored it as not relevant. The criteria for evidence inclusion and exclusion used for AQ2 are the same as reported in the first and second report for this mandate (EFSA, 2025, 2026) and all previous considerations apply. AQ3. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to animals? The performance of AI as a second reviewer had a low disagreement rate between human and AI assessments (34/597 records, 4.7%). Discrepancies were evaluated by an additional EFSA reviewer, and in 30/34 (88.2% of the cases the human assessment was confirmed). Among the conflicting cases, 6 (6/34, 17.7%) were identified as false negatives, and four of these was later deemed relevant following full‐text screening. As shown in Table 5, most conflicts were associated with the criterion ‘Outcome/conclusions’, whereas a smaller number related to organism, scope/intended use or publication type. The outcome indicates that AI performed well as a second reviewer for AQ3, but the possibility of false negatives still requires the resolution by a third reviewer. As the training dataset increases, the sensitivity is expected to improve. TABLE 5Conflict resolution between human and AI reviewer for AQ3.Criteria for exclusionFalse positives a False negatives b Agreement with AI reviewerAgreement with human reviewerAgreement with AI reviewerAgreement with human reviewerOrganism1Outcome/conclusions2345Publication type11Scope/intended use1Total: 24Total: 10 Note: Each record could have had more than one criteria for exclusion. a ‘False positives’ are the records which the human reviewer scored as not relevant, while AI scored it as relevant. b ‘False negatives’ are the records which the human reviewer scored as relevant, while AI scored it as not relevant. Conflict resolution between human and AI reviewer for AQ3. Note: Each record could have had more than one criteria for exclusion. ‘False positives’ are the records which the human reviewer scored as not relevant, while AI scored it as relevant. ‘False negatives’ are the records which the human reviewer scored as relevant, while AI scored it as not relevant. The criteria for evidence inclusion and exclusion used for AQ1 are the same as reported in the first and second report for this mandate (EFSA, 2025, 2026) and all previous considerations apply. ToR3. Extract and summarise relevant data and evidence and assess whether any new data and evidence may have implications for EFSA relevant scientific opinions AQ1. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to plants? Three publications were critically discussed further with experts to assess whether they provide potentially relevant evidence which may have an impact on the previous EFSA conclusions (Table 1). These studies are reported in Annex B and regard the use of AI as a supporting tool in experimental design of NGT plants (Juhas et al., 2025), CRISPR‐Cas9 genome editing in rice (Qin et al., 2025) and chimerism observed in Cas9‐free segregants of gene‐edited alfalfa (Santoro et al., 2025). Juhas et al. (2025) examine the potential implications of combining AI with NGTs, focusing on whether AI‐assisted methods could support the development of plant traits that fall within regulatory categories not requiring environmental risk assessment (NGT1). As a proof of concept, the authors describe the in‐silico design, using a generative AI tool, of a maize plant with increased expression of an endogenous serine protease inhibitor (SPI), conferring insecticidal properties. The approach and the use of AI in the study appear primarily to support the identification and conceptual combination of strategies already described in the scientific literature, rather than demonstrating the development of a novel tool or design. In particular, the proposed genetic modifications target regulatory regions, such as cis‐regulatory regions (CREs) and upstream open reading frames (uORFs), which play a key role in controlling gene expression. Genetic variation affecting these regulatory elements is widespread in plant genomes and represents an important source of phenotypic diversity. Modifications in these regulatory elements already occur naturally in plant genomes and have been subject to both natural and human selection during domestication and breeding (Cui et al., 2023; Li & Schmitz, 2025; Wang et al., 2024). Qin et al. (2025) describe CRISPR‐Cas9‐mediated genome editing in rice and report the unexpected detection of the hpt selectable marker in progeny of plants initially considered transgene‐free. The authors attribute this observation to mechanisms such as undetected low‐frequency or partial T‐DNA insertions, tissue mosaicism or outcrossing events. These findings highlight limitations in screening approaches for confirming the absence of transgenes and detecting mutations, particularly in early‐generation material, where comprehensive analyses (e.g. whole‐genome sequencing or Southern blotting) were not performed. The study concerns pre‐breeding material, where chimerism and somatic mosaicism are expected may occur as a result of transformation and regeneration processes. These processes can lead to heterogeneous cell populations and may affect the consistency of detection of transgenic elements and mutations across generations. Comparable observations are reported by Santoro et al. (2025) in tetraploid alfalfa. Deep amplicon sequencing identified multiple low‐frequency edited alleles, possibly indicating continued activity of the genome‐editing machinery following in vitro regeneration. Importantly, no new edited alleles were detected after sexual reproduction, suggesting that these effects are associated with post‐editing propagation dynamics rather than with the CRISPR/Cas9 approach per se. Chimerism was also observed at low frequency in Cas9‐free segregants, which the authors hypothesise could result from residual or transient activity of the editing machinery (e.g. persistence of Cas9 protein or maternal effects), although this remains speculative and has not been experimentally demonstrated. Nevertheless, these phenomena are consistent with processes occurring during the early stages of plant regeneration and pre‐breeding, including in conventional breeding (EFSA GMO Panel, 2021b). The chimeric presence of CRISPR‐Cas components or transgenic sequences should be appropriately assessed across the plant to demonstrate their absence in the final material. This aspect is relevant to both sexually and vegetatively propagated crops and primarily concerns early stages of plant development and pre‐breeding material. Chimerism in plants is a well‐documented biological phenomenon that occurs both naturally and as a result of human selection, particularly in ornamental plant breeding, where it contributes to desirable variegated phenotypes (Park et al., 2022). Chimeric tissues may also arise during plant regeneration processes, including those involving tissue culture and genome‐editing techniques. In such contexts, independently edited and non‐edited cell lineages may persist within the same individual, leading to mosaic or sectorial genetic patterns (Faize et al., 2010). Although chimerism may affect the genetic uniformity, stability and representativeness of the plant material, its occurrence does not, per se, indicate the presence of novel hazards. None of the studies discussed provided new evidence that changes the previous EFSA conclusions. AQ2. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to microorganisms? One publication was discussed further with experts to assess whether it provides potentially relevant evidence, which may have an impact on previous EFSA conclusions and recommendations (Table 1). This study reported in Annex B, is authored by Overton et al. (2026) and concerns the mutational burden imposed by a CRISPR‐Cas9 gene‐drive element. Overton et al., 2026 investigate how the presence of a CRISPR‐Cas9 gene drive (CCGD) or an untargeted Cas enzyme by itself in the genome of Saccharomyces cerevisiae could affect the rates of loss of heterozygosity (LOH) and de novo mutations, potentially altering long‐term evolutionary trajectories of treated populations. The authors also try to quantify the magnitude of these effects and their evolutionary significance. The authors performed mutation accumulation experiments and tested three different constructs, namely Wildtype (W), Cas9 (C) and Drive (D). They concluded that the 30% increase in LOH observed was comparable among the three constructs and based on literature searches, stated that the possibility of a 30% increase in the LOH event rate due to CCGD carriage was quite modest compared with the natural variation in LOH rates across different strains of S. cerevisiae. The low rate of off‐target mutations may be at least partly contributed to the well‐designed gRNA, which helps minimise off‐target mutations and is expected to become common practice. In its scientific opinion, the EFSA GMO Panel (2024) addresses the issue of mutations and states that the same mutations (and related hazards) can often be introduced by different techniques (conventional mutagenesis, EGTs and NGTs). Therefore, the assessment should be on the genotype and phenotype of the resulting microorganism and not on the technique used. Specifically for engineered gene‐drives or similar technologies designed to bias, and therefore speed up, the transmission of certain genetic elements in a target population, EFSA states that additional guidance is recommended to be developed. The experimental set‐up and considerations by Overton et al. (2026) could inform its development in the future. Overall, this study does not present any information that would change the recommendations and conclusions of the existing EFSA opinions. AQ3. Is there new evidence emerging from the identified studies that may have implications for relevant EFSA scientific opinions on NGTs applied to animals? Four studies were discussed further with experts to assess whether the results reported therein potentially provide evidence that could change previous EFSA conclusions (Table 1). The studies are reported in Annex B and regard chromatin perturbations observed after genome editing (Zhu et al., 2026); RNA off‐target editing and associated cytotoxicity induced by adenine base editors (Zhang et al., 2026); disabilities associated to the expression of a cytochrome P450 reductase (Cypor)‐targeting short hairpin RNA (Martinez et al., 2025) and integration of concatemers of the AAV vector in the Cas‐9 induced DNA breaks (Martinez et al., 2025; Skov et al., 2025). Zhu et al. (2026) aim at assessing how genome editing impacts chromatin regulatory networks. The publication reports that editing events in non‐coding DNA could cause chromatin perturbation that extends along DNA, with the span varying among different cell types. Genome editing significantly activated the p53 pathway which modulates gene expression, but it would not be the only element leading to transcriptomic alterations. Distal genes, closer in 3D (three dimensional) space to the editing site, appeared to be more affected than those further away. This study demonstrates that DNA cleavage triggers a surge in chromatin accessibility that propagates far beyond the immediate edit site. To preserve cell identity and minimise these architectural disruptions, the authors proposed mitigation strategies, such as Alternative Editing Systems or optimised sgRNA design workflow. The observed effects are not unexpected, P53 and similar pathways are likely to be activated in consequence of DNA damage such as that caused by double stranded breaks. The findings in this publication are relevant for NGT applications on cell cultures; the reported chromatin perturbations are transient and limited to edited cells in the presence of the reagents. These effects, even if they occurred in edited cells of a zygote, would no longer be present by the time the edited animal is born. Therefore, commercialised animals would not be affected by the negative effects reported in this publication. The study by Zhang et al. (2026) investigates RNA off target (OT) editing induced by adenine base editors (ABEs), specifically the TadA deaminase component, and its link to toxicity, while developing tools and engineered variants to minimise RNA editing and improve safety. Widespread RNA OT editing mediated by the TadA deaminase is the primary driver of observed cytotoxicity and in vivo toxicity, including impaired cell viability and lentiviral defects. The authors developed a fluorescent RNA inosine sensor (MINOS) enabling rapid, sensitive and high‐throughput quantification of RNA editing activity. Several TadA mutants with minimised RNA editing activity were identified. These variants generated functional ABEs with substantially improved safety profiles, including restoration of cell viability, elimination of in vivo toxicity and significantly reduced RNA off‐target effects. The cytotoxic effects of ABE are known (EFSA, 2026). The reported adverse effects are due to the high dose of editor used. In general, effects of ABE on cellular RNA are transient and only present as long as the editor is active. This publication provides solutions towards a better editing system and shows that this is an evolving field. The reported adverse effects might be relevant for gene therapy applications but would not affect gene‐edited animals meant to be commercialised. The publication by Martinez et al. (2025) describes ways to improve the homology‐directed repair (HDR)‐mediated gene insertion for the treatment of murine phenylalanine hydroxylase (PAH) deficiency. Simultaneous pharmacologic inhibition of the competing non‐homologous end joining (NHEJ) and microhomology‐mediated end joining (MMEJ) pathways synergistically promotes the higher‐fidelity HDR pathway. The publication reports severe motor disability, motor dysfunction in mice treated with high‐dose AAV8 vectors carrying a cytochrome P450 reductase (Cypor)‐targeting short hairpin RNA (shRNA), meant to provide a selective cell growth advantage. The described negative effects are not due to the gene editing technique, but rather associated to the expression of shRNA cassette, which has already been reported in other publications cited in this paper (Ehlert et al., 2010; Martin et al., 2011; McBride et al., 2008). The publication also reports integration of concatemers of the entire AAV genome into the Cas9‐induced DSB. Undesired integration of AAV vectors is known, it is dose dependent and seems to be especially evident with the combination of AAV vector and CRISPR‐Cas. NGT animals to be commercialised should always checked for the absence of undesired integration of the vector used during gene editing and other exogenous sequences. Skov et al. (2025) describe a therapeutic strategy for GATA2 deficiency based on CRISPR‐Cas9‐based gene correction, utilising recombinant adeno‐associated virus serotype 6 (rAAV6) as a template for HDR. High cytotoxicity in haematopoietic stem and progenitor cells and a high degree of large aberrations at the on‐target site in haematopoietic stem and progenitor cells (HSPCs) due to the integration of AAV concatemers, are reported. Both toxicity and undesired integration of AAV vector are known phenomena and are associated with the high doses of the reagents used. While these effects are important in gene therapy applications, they are not relevant to gene‐edited animals, because the dose of vector or Cas9 used can be adjusted and potential adverse effects are transient and would disappear during the subsequent cell divisions that result in the edited animal. None of the studies discussed provided new evidence that changes the previous EFSA conclusions. CONCLUSIONS The outcome of the screening for AQ1–3 demonstrates that the search strategy (ToR1) is both sensitive and specific. It reliably identifies publications of relevance while limiting the inclusion of irrelevant publications. The outcome of the screening for AQ1–3 also confirms that the inclusion and exclusion criteria (ToR2) are specific and appropriate for the mandate. The application of a conservative approach to studies of uncertain relevance provides additional assurance that no significant publications were overlooked during the initial screening phase. The implementation of AI to act as reviewer for title/abstract screening was successful, leading to more efficient screening with low disagreement rate between the human and AI reviewer. Disagreements were resolved by a third human reviewer. In line with the EFSA protocol and the assessment questions, EFSA concluded that none of the studies identified through the literature search introduces hazards or risks not previously considered in EFSA scientific opinions (ToR3). RECOMMENDATIONS FOR EFSA'S FUTURE WORK EFSA could continue to systematically assess the suitability of the search strings used in the literature screening protocol as NGTs evolve and their applications in the agri‐food and environmental domain relevant to the past EFSA opinions expand. Any modifications to the protocol would be documented and justified. EFSA would also ensure consistency in the screening process through internal calibration exercises between reviewers. Across the three scientific reports prepared to date, more than 8000 studies have been screened at the title/abstract level. Building on the successful implementation of the DistillerSR AI‐assisted title/abstract screening, EFSA could continue to use these screened studies as training dataset and maintain the AI‐supported second‐reviewer approach in subsequent reporting cycles. Despite the large number of screened publications, the number of studies deemed relevant at the full‐text level remains limited and these studies are characterised by heterogeneous experimental designs. Consequently, the development of structured critical appraisal tools is currently considered to provide limited added value. EFSA would therefore continue to apply a qualitative approach to evidence extraction and appraisal. To structure and catalogue the assessed studies, EFSA has developed a searchable archive, currently as an Excel‐based database (Annex B) which includes all the publications screened at the full text level. This database contains the appraisal outcomes for each study and enables filtering and retrieval of studies according to key criteria, such as EFSA report number, organism group (e.g. plants, microorganisms or animals), species, experimental techniques. Looking ahead, further work may focus on improving the presentation and accessibility of the data through the development of more visual and user‐friendly tools such as Power BI dashboards. In the longer term, this approach could support the creation of a comprehensive and continuously updated repository of publications relevant to NGTs in the context of food, feed and environmental risk assessment. Such a resource would represent a valuable asset for future EFSA activities, including the development of scientific guidance and evidence‐based decision‐making. ABBREVIATIONS adeno‐associated virus adenine base editor artificial intelligence assessment question CRISPR‐Cas9 gene drive cis‐regulatory element clustered regularly interspaced short palindromic repeats established genomic technique genetically modified organism homology‐directed repair haematopoietic stem and progenitor cells loss of heterozygosity molecular characterisation mitochondrial inner‐membrane organising system microhomology‐mediated end joining new genomic technique non‐homologous end joining off‐target phenylalanine hydroxylase Preferred Reporting Items for Systematic reviews and Meta‐Analyses short hairpin RNA serine protease inhibitor subquestion transcription activator‐like effector nucleases Term of Reference upstream open reading frame Working group cross‐cutting REQUESTOR European Commission QUESTION NUMBER EFSA‐Q‐2026‐00084 COPYRIGHT FOR NON‐EFSA CONTENT EFSA may include images or other content for which it does not hold copyright. In such cases, EFSA indicates the copyright holder and users should seek permission to reproduce the content from the original source. Supporting information Annex A: Protocol (incl. its appendices) Annex B: Repository of studies thar underwent full‐text screening