Google just proved it can build artificial intelligence sophisticated enough to read sign language and translate it into text in real time — yet it still insists it has no choice but to censor your posts with blunt, unaccountable algorithms.
At its Made by Google 2026 event this week, the company unveiled a new AI model called SL2T, built by Google DeepMind, that uses the Pixel 11's camera to recognize American Sign Language and convert it to English on the fly. The model was trained on more than 100,000 hours of data across over 50 sign languages, according to Digital Trends. It processes body-movement coordinates on-device — not raw video — then sends only those coordinates to Google's servers for translation before discarding the camera feed. Google says it even tested for left-handed signers and one-handed signing. The company claims it scored higher on a standard sign language translation benchmark than any previous system.
This is a genuine technical achievement. An estimated 70 million deaf and hard-of-hearing people worldwide use sign language, and Google DeepMind says this marks the first time sign language AI at this scale has reached a consumer product. Digital Trends noted that translating sign language is "fundamentally different" from translating speech, since sign languages have their own independent grammar and vocabulary rather than word-for-word mapping.
So let's take stock. Google can train a model on 100,000 hours of nuanced, visual, grammatically distinct human communication, run privacy-protecting on-device processing, and ship it in a phone. But when it comes to content moderation — deciding whether a post violates community guidelines — the company claims the task is too complex, too nuanced, too prone to error to do anything other than automated takedowns and opaque appeals processes. The technology exists to serve everyone, except those who dissent from the establishment narrative.
TechCrunch, for its part, buried the sign language feature inside a broader roundup of Gemini updates, framing it as an "expansion of Live Transcribe" rather than the standalone breakthrough Digital Trends described. 9to5Google didn't mention it at all, devoting its coverage to the new "Extreme Charging Mode" — a feature that lets the phone get hotter and suspend background tasks to charge slightly faster within the same 30-45 watt caps. Priorities.
The Pixel 11 lineup itself starts at $899, up $100 from the Pixel 10, with the Pro starting at $1,099. TechCrunch reported that Google doubled base storage to 256GB but dropped the cheaper 128GB option, and cited the ongoing RAM supply shortage as part of the price hike. The Pro Fold runs $1,899. Google also launched a $29 Pixel Tag to compete with Apple's AirTag.
None of this is trivial. The sign-to-text feature could genuinely improve lives. But Google's selective competence tells you everything about where its priorities lie. When the goal is expanding the user base and generating good press, the engineering problems are solvable. When the goal is fairly moderating political speech, suddenly the algorithms are just too imperfect. The capability gap isn't technical. It's intentional.
Whether SL2T holds up outside benchmark conditions — poor lighting, fast casual conversation — remains an open question, as Digital Trends noted. Whether Google will ever apply the same ingenuity to protecting free expression is another one entirely.
