Pangram releases Pangram Image 1.0, which it says is state-of-the-art in AI image and AI video detection, and the smaller Image Flash 1.0 for AI video detection

AI-generated images and videos are now detectable with significantly higher accuracy, impacting content authenticity.

Pangram has released Pangram Image 1.0, a new AI image and video detection model. The company claims it achieves state-of-the-art performance, reporting 99.96% accuracy on a test set of 1,200 synthetic images generated by 12 frontier AI image generators. The model is also robust to common image manipulations like compression, edits, and low resolutions, with detection accuracy maintained down to 256x256 pixels. Additionally, Pangram introduced Pangram Image Flash 1.0, a faster model optimized for AI video detection, which reportedly scores 98.7% on the DF26 benchmark, a significant improvement over existing detectors. Flash 1.0 analyzes videos frame by frame and is available via API and as a Chrome extension, alongside Pangram Image 1.0.

Pangram's new models represent a substantial leap in AI content detection capabilities, particularly for video. The reported 98.7% accuracy on the DF26 benchmark for Pangram Image Flash 1.0 suggests a significant market shift towards tools that can reliably distinguish AI-generated video from real footage, a capability previously described as humans being barely better than random at identifying. This improved video detection, coupled with the high accuracy for image detection, indicates that the arms race between AI generation and detection is escalating, with detection tools now achieving near-perfect results in specific benchmarks.

The release of robust, high-accuracy AI detection tools like Pangram Image 1.0 and Flash 1.0 will likely influence how platforms and users approach content verification. With API access and a Chrome extension, these detectors are becoming more accessible, potentially leading to widespread adoption for flagging AI-generated content across social media and other digital platforms. This improved detection capability could drive greater demand for AI content authenticity solutions and push AI generation model developers to focus more on evasion techniques.