WebSpeech: SpeechSynthesisUtterance does not strip XML from SSML text; speaks literal markup. From https://dvcs.w3.org/hg/speech-api/raw-file/9a0075d25326/speechapi.html#utterance-attributes "[the text] may be either plain text or a complete, well-formed SSML document. [SSML] For speech synthesis engines that do not support SSML, or only support certain tags, the user agent or speech engine must strip away the tags they do not support and speak the text." XML markup examples in the SSML spec: http://www.w3.org/TR/speech-synthesis/
<rdar://problem/15129950>
Related to the larger SSML TLF bug 158387
Mark Hakkinen from Education Testing Service (ETS) mentioned they are having to work around this bug now.
Relevant spec issue: https://github.com/w3c/speech-api/issues/37