{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreibcwjv7c2emp5oedmibvzzxhpjcqqfc4h5khtjpczk2k6l3ylzhpi",
"uri": "at://did:plc:mxzzpugn7bprjjrszwkbez3u/app.bsky.feed.post/3mjp4lh3k2be2"
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreidd7nbakg2udvxu2ojsng3sttjwgv6ccnaj2mtuyflvnebkw7mjte"
},
"mimeType": "image/jpeg",
"size": 59831
},
"path": "/news/2026-04-ai-image-generators-safety-hidden.html",
"publishedAt": "2026-04-16T16:00:04.000Z",
"site": "https://techxplore.com",
"tags": [
"Machine learning & AI"
],
"textContent": "Generative AI models can be prompted with just a few words to insert offensive or discriminatory text messages into images. Aditya Kumar from the SPRINT-ML Lab at the CISPA Helmholtz Center for Information Security is investigating how such outputs can be reliably prevented. To address this, he developed ToxicBench, a test dataset that evaluates how well image-generating AI systems handle offensive inputs. He also created a fine-tuning strategy to adapt the models accordingly.",
"title": "AI image generators get a new safety test for hidden toxic text in memes"
}