With more realistic images than before, GPT Image 1.5 fares reasonably well against Google's Nano Banana Pro in my testing.
An 'automated attacker' mimics the actions of human hackers to test the browser's defenses against prompt injection attacks. But there's a catch.
As usual, you can take advantage of custom instructions to guide ChatGPT's personality in a direction you like, especially when the presets don't give you those options. For example, if you'd like ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results