A new tool called Pangram is trying to change how we spot AI-written text. Developed by a team of engineers with backgrounds at Tesla and Google, it claims to be the best at catching AI content. Pangram says it can identify AI-generated writing with over 99% accuracy in multiple languages like English, Spanish, and Arabic. That sounds impressive, but real-world use can be more complicated.

What Makes Pangram Different?

Most AI detectors focus on finding patterns that are typical of machine writing. Pangram tries to go a step further by spotting hybrid content—texts that mix human and AI writing. This is important because many teachers, journalists, and lawyers struggle to tell the difference. If Pangram works well, it could help these professionals see through the fake stuff and trust what they read.

But detecting hybrid content isn’t easy. Human writers often produce text that can seem a little “formulaic” or robotic, especially when they rush or try to meet a deadline. Critics worry that detectors like Pangram might flag genuine human work as AI, raising questions about fairness. Schools and workplaces have already faced issues with false alarms, and this new tech might make that worse if not used carefully.

Beyond Text: Protecting Identity and Legal Battles

Detection isn’t just about written essays anymore. Platforms like YouTube are now using similar tools to protect faces and voices from deepfake videos. Recently, YouTube announced it will expand this technology to its Partner Program, allowing creators to report fake videos that use AI to mimic their appearance or voice. This shows that AI detection is becoming a tool for safeguarding personal identity, not just filtering content.

Legal issues are also heating up. A group of YouTubers has accused OpenAI of using their transcribed videos without permission to train AI models. This indicates that AI detection and content ownership are now intertwined. Companies are starting to see detection as a way to defend their rights and control over their digital likeness.

The Future of AI Detection and Its Challenges

Pangram’s developers are confident their tool can restore trust in online content. If it truly works as promised, it could help differentiate real from fake and bring some clarity to a confusing landscape. Still, the history of AI detectors isn’t perfect. Past tools have had false alarms, mistakenly flagging human creativity as AI-generated. This has caused frustration and even accusations of unfair censorship.

As the technology matures, it might improve. But until then, users should stay cautious. Detection tools are helpful, but they’re not foolproof. In the ongoing game of cat and mouse, AI creators and detectors will keep evolving. The stakes are high—not just for tech companies but for society as a whole. Protecting truth, privacy, and ownership online is becoming more critical than ever.

In the end, tools like Pangram could be a step toward safer, more trustworthy digital spaces. But trust must be earned, and the technology will need to prove itself in real-world situations. For now, it’s a watch-and-see game with big implications for all of us.

Inspired by

• https://ai2people.com/pangram-wants-to-sniff-out-ai-text-but-can-it-really-catch-the-bots/