Google's solution to images altered with their AI: just use their AI to determine if they're fake.

There's a little problem though: people who assume it's real won't get told. Not even through their own AI. I downloaded the fake protest picture and asked, using the persona of someone naive.

Then I asked if it was real. Only then did it bother to check:

So Google will only fulfill their promise if you yourself act as a skeptic at all times. I love the erosion of trust online, as pushed by major tech companies.