News

Deliberate circumvention of risk assessment for NGT plants with ChatGPT

In a joint "Proof-of-Concept" experiment by the Aurelia Foundation, the Testbiotech association and the Save Our Seeds (SOS) initiative, it was demonstrated that ChatGPT can be used to develop NGT plants that meet all the criteria of the current EU legislative proposal on NGT regulation in order to be released without risk assessment – even though they are in fact fraught with risk.    

Experiment reveals weaknesses in the criteria of the EU legislative proposal

According to the current legislative proposal by the EU Commission, so-called "NGT 1 plants" from new genetic engineering, which have been modified at a maximum of 20 sites in the genome and do not contain any foreign genetic material, could in future be released without risk assessment and labelling. In the experiment, the publicly accessible version of the AI bot ChatGPT 4o designed an insecticidal NGT1 maize plant based on precisely these criteria. This shows that, on the basis of the legislative proposal, risk-fraught NGT plants can be specifically developed, whose environmental compatibility would then not have to be tested before release.

Case-by-case assessment of all NGT plants urgently needed

Artificial Intelligence that is accessible to everyone opens up new possibilities for creating risky NGT plants very quickly and easily, which can easily circumvent the criteria for a risk assessment. This once again demonstrates the need for case-by-case assessment, even for NGT1 plants, in order to comply with the precautionary principle.

Testbiotech: ChatGPT-designed insecticidal NGT plants

Aurelia Foundation: ChatGTP builds insecticidal genetically engineered plant (German)

Start of trilogue negotiations on the regulation of new genetic engineering