The development of Proto-Germanic ai in the North Germanic languages is very complex. It depends on a number of factors (stressed/unstressed, followed by w, r/R, h etc.) - see e.g. [Sandøy 2017][1]

>**PGmc. ai > Proto-Norse æi** ([Fulk 2018:][2] 70), which further developed into:

 - ǣ (before w) ([Heusler 1921: 28][3]);
 - ā (before r, h) (this ā further developed into ē in the i-umlaut environment);
 - ei (in OWN) or  ē (OEN) ([Versloot 2017:][4] 285);

In your post, you ask about frǽ (and its variant frjó; [Zoega 1926][5] says that fræ was "sometimes in old MSS. spelt freo or frjó (q.v.), but less rightly").

As you can easily notice, we have *w* after the diphthong in PGmc, which means PGmc ai developed into æ in Old Norse (see above), cf. Fulk 2018: 70

"In  addition,  Proto-Norse  æi developed to ON æ (i.e., /æː/) before w (which might be lost, §6.14), as in *aiwīn- > OIcel. ævi ‘age’ (cf. Go. aíws < *aiwaz) and hræ ‘corpse’ (cf. Go. hraíw)."

*Note:* the acute and the macron are used interchangeably here, to indicate length.

  [1]: https://www.degruyter.com/view/book/9783110197068/10.1515/9783110197068-084.xml
  [2]: https://www.jbe-platform.com/content/books/9789027263131
  [3]: https://archive.org/details/altislndischesel00heus/page/28/mode/2up
  [4]: https://www.degruyter.com/document/doi/10.1515/flih-2017-0010/html
  [5]: https://onp.ku.dk/onp/onp.php?o24213#