Misinformation and Model Poisoning in LLMs

Public preview

Want everyone on your software team to learn this? Free 14-day trial.

Train my team

Bias can emerge from biased training data, incomplete representation, flawed algorithms, or biased human feedback during development. If your model works with data on humans, you need to be careful to test and allow for bias in your model.

An older man in a suit and a younger female journalist
GenAI image tools like Midjourney show women for inputs including non-specialised job titles such as journalist, but show older men for specialised roles such as news analyst.