Former Google Engineers Built a Tool to Shield Kids from Harmful Texts

Former Google engineers have launched a groundbreaking AI-driven platform designed to shield children from psychologically damaging online texts—ranging from self-harm encouragement to sexual grooming narratives. This initiative is a direct evolution of Google’s landmark 2013 child pornography takedown operation, shifting focus from visual content to the insidious, text-based threats now proliferating across forums, social media, and messaging apps.

The Hidden Danger of Harmful Texts

In the late 2010s, a new category of digital threats emerged: so-called 'toxic texts'—written content disguised as innocent or poetic messages, but designed to manipulate vulnerable children into self-harm, isolation, or sexual exploitation. Traditional keyword filters failed to catch phrases like 'no one cares if I disappear' or 'the pain will end if you stop breathing,' which use emotional nuance and coded language. Recognizing this gap, a team of seven ex-Google AI researchers, a child psychologist, and a child rights attorney developed SafeText: a contextual language model trained on thousands of real-world examples of harmful child-directed texts. Unlike conventional filters, SafeText analyzes tone, emotional trajectory, and linguistic patterns to identify threats before they escalate.

From Google’s Legacy to an Independent Mission

While Google’s 2013 child pornography operation earned global acclaim, internal priorities later shifted away from proactive content moderation. Disillusioned by corporate inertia, the former engineers launched SafeText independently in 2022. The system now operates as a browser extension and API, integrated into school networks, parental control apps, and youth-focused platforms. Initial trials showed a 94% detection rate for harmful texts, with 87% of parents reporting reduced exposure in their children’s online activity. In 2024, SafeText partnered with a Turkish educational foundation to deploy the tool in 150 schools across Turkey, training teachers to use it as a real-time safeguard.

Google has not officially endorsed the project, but its founders argue it embodies the company’s original mission: 'to organize the world’s information and make it universally accessible and useful.' 'We didn’t leave Google to build another search engine,' says the project lead. 'We left to protect the next generation from the invisible wounds of the digital age.' With funding from nonprofit tech foundations and growing adoption in Europe and North America, SafeText is poised to become the global standard for child-safe text filtering—proving that innovation doesn’t always come from the boardroom, but from those who refuse to look away.