Literary translation is generally considered more of an art compared to general translation. I am aware of 

 where a non-fiction book (fittingly, Daniel Susskind's “A World Without Work. Technology, Automation, and How We Should Respond”) was "translated by an AI"; however, in practice it seems to be what is euphemistically called "machine translation post-editing", which is basically a machine translation which is (sometimes heavily) edited by a translator, and then further edited by an editor. Thus, it was neither a real "translation" nor a really literary text.

This market resolves once three major works of literature have been published by major publishers on major Western book markets in translations produced entirely by software.

Major works of literature might be defined e.g. as winning a major award (think Pulitzer, Booker or Prix Goncourt) or being a major bestseller. I think re-translaitons of classic works would not qualify. Non-fiction works will not qualify.

Major publishers are even harder to define and will be evaluated subjectively; my intent is to exclude publishers which might not care enough about their reputation and be satisfied with a lower-quality translation, or which might do it 

Major Western book markets would be e.g. US/UK, France or Germany, but nor ones like Latvia (too small) or Japan (which I might know enough about to evaluate claims properly).

Translations produced entirely by software means that no human translator or editor should be involved in the translation and publishing process. A proofreader would be acceptable.

After the conditions are met, I would wait for 6 months before resolving the market in order to make sure no cheating (such as human involvement) becomes public knowledge.