John Mueller @johnmu.com · 02/04/2026Interestingly, it doesn't throw off the newer Gemini models that I started using for this. Nice work, Deepmind folks, and sorry to everyone :-). 250 5
Martino Mosna @martinomosna.com · 02/04/2026Basic prompt injection techniques are taken care of by the major models, luckily. But there's always room for harm, after all LLMs are probabilistic machines and the guardrails are never 100% foolproof. 010 1
