No, as far as I know DeepMind's AlphaZero chess program has not been publicly released in any form, and I don't know if they plan to do so at any point. The only games of AlphaZero vs Stockfish that have been publicly showcased are the 10 games that can be found either in their pre-print paper, on their website or the chess news outlets such as chess24.
I guess the comment you're referring to is the one made during the press conference by Alexander Grischuk after his first game vs Liren Ding, where they played the Anti-Moscow variation of the semi-Slav Defense, an opening that has not occurred in any of the published sample games of AlphaZero vs Stockfish. So for all we know, as ETD has also pointed out in comments, Grishuk is jokingly labeling probably his
16.a4 novelty (or possibly the whole sacrificial variation after
Nxf7) as an idea liked by AlphaZero but disliked by Stockfish. This is also pointed out in the chess24 report of the 4th round.
16.a4 in this variation where the main move is usually
16.Bg4 (which became famous after the Topalov vs Kramnik game in 2008):
[title "Semi-Slav Defense: Anti-Moscow variation 16.a4"]
1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Nc3 c6 5. Bg5 h6 6. Bh4 dxc4 7. e4 g5 8. Bg3 b5 9. Be2 Bb7 10. O-O Nbd7 11. Ne5 Bg7 12. Nxf7 Kxf7 13. e5 Nd5 14. Ne4 Qb6 15. Nd6+ Ke7 16. a4 (16. Bg4)