Generative Adversarial Network (GAN)

Last Updated: July 29, 2026 | By Mihail Sebastian | AI Dictionary

A deep learning framework in which two neural networks compete: one generates synthetic data, the other judges it, until the fakes pass for real.

What is a Generative Adversarial Network (GAN)?

A generative adversarial network (GAN) is a deep learning model built from two neural networks locked in competition: a generator that produces synthetic data and a discriminator that tries to tell fake from real.

Ian Goodfellow and colleagues introduced GANs in 2014. The design’s insight is that competition itself is a training signal: the generator improves precisely because the discriminator keeps catching it.

How a GAN Works

The generator starts from random noise and outputs a candidate sample, at first meaningless static. The discriminator receives a mix of real training examples and generated fakes, and labels each one real or fake.

Both networks learn from the verdicts. The generator adjusts to fool the discriminator; the discriminator adjusts to catch the improved fakes. Training succeeds when the discriminator does no better than a coin flip, because at that point generated samples are statistically indistinguishable from real ones.

Applications of GANs

Example of a GAN

Train a GAN on a large collection of photographed human faces. Early in training the generator emits noise and the discriminator rejects everything. Round by round, the generator learns the geometry of faces: first blobs with eyes, then coherent features, eventually portraits with pores and stray hairs.

The finished generator produces a photorealistic portrait of a person who has never existed. The same capability powers deepfake impersonation, which is why the EU AI Act requires deepfakes to be disclosed as artificially generated.

Related AI terms: Generative AI · Variational Autoencoder · Synthetic Data · Data Augmentation

Did you like the Generative Adversarial Network (GAN) gist?

Learn about 250+ need-to-know artificial intelligence terms in the AI Dictionary.

Mihail Sebastian — Writes about AI governance, regulation, and the technology behind them. Placeholder bio — replace with a real credential line. About

Read the Governor's Letter

Stay ahead with Governor's Letter, the newsletter delivering expert insights, AI updates, and curated knowledge directly to your inbox.

By subscribing to the Governor's Letter, you consent to receive emails from AI Guv.
We respect your privacy - read our Privacy Policy to learn how we protect your information.

Browse All AI Terms A–Z

Every term in the dictionary, in alphabetical order. Jump to a letter or scroll the full list.

A

B

C

D

E

F

G

H

I

J

K

L

M

N

O

P

Q

R

S

T

U

V

W

X

Y

Z