Pangram привлёк $9 млн на детекцию AI-контента и выпустил модель Pangram 4

Startup Pangram raised $9 million on July 29, 2026 to scale its software for detecting content generated by artificial intelligence. Alongside the round, the company released a new text detection model, Pangram 4, and its first AI image detector in research preview status.

What Pangram released

Pangram unveiled two products at once. The main one is an updated AI text detection model, Pangram 4, which determines whether a piece of text was written by a human or a neural network. The second is a detector for AI-generated images, currently available in research preview format — an early research version. Both tools solve the same task — distinguishing human-created content from machine-generated content — but work with different types of data.

• Round: $9 million, announced July 29, 2026
• New text model — Pangram 4
• AI image detector — in research preview status
• Purpose of the funds — scaling software for AI content detection
• News source — TechCrunch

How much Pangram raised

Pangram raised $9 million — these funds, according to TechCrunch, will go toward scaling software for detecting AI-generated content. The round coincided with the release of two models at once, which shows the startup's bet on expanding from text to images.

For the AI detection market, $9 million is a moderate but notable check: it allows the company to hire a team, increase computing power for training detectors, and cover new content formats.

Why detect AI content

Detecting AI content is necessary because texts and images created by neural networks are increasingly difficult to distinguish from human-made ones. Since late 2022, following the mass spread of generative models, the volume of such content on the internet has sharply increased — from student papers and articles to images in social media and advertising.

The issue is especially acute in education: teachers need to know whether a student actually wrote an essay themselves, and plagiarism-checking services are adding AI detection to their products. Pangram covers both formats: first text, now images too.

How AI text detection works

Detectors like Pangram 4 analyze statistical features of text: how predictably phrases are constructed, how words and expressions are distributed. Models trained on "human vs. neural network" pairs look for patterns characteristic of machine generation. Accuracy depends on the volume and diversity of training data — which is why newer versions are usually more accurate than previous ones.

The image detector works differently: it looks for generation artifacts — atypical textures, inconsistencies in details, traces of specific neural networks. Since Pangram's image product is still in research preview status, the company will likely continue collecting data and fine-tuning it before a full release.

Why detectors aren't always trusted

AI text detectors are criticized for false positives. According to a 2023 Stanford study, early detectors more often mistakenly flagged texts written by non-native English speakers as "machine-generated." That's exactly why the accuracy of new models like Pangram 4 is a key parameter: a single mistake can cost a student an accusation of cheating, and an author their reputation.

"AI content is flooding the internet — and

Pangram raised $9 million to detect it," — as stated in the TechCrunch article.

What this means

As generative models become cheaper and more numerous, demand for tools that distinguish machine content from human content is growing. Pangram's round and the launch of the image detector signal that AI detection is becoming a separate market rather than a side feature. For authors, teachers, and platforms, this means that checking the origin of content is gradually becoming a standard procedure rather than an exotic one.