Even if they’re “trying”, it’s moot if the result isn’t clearly more ethical, and with the proliferation of stolen imagery on their stock image service (which they use to train their models), the ethics of their models are very much not clear.

If I saw news of a huge purge of stolen content on their stock image service with continued periodic purges afterwards (and subsequent retraining of their models to exclude said content), I might take the claim more seriously.