"AI is perfectly safe and has no bias, risks, or safety concerns whatsoever", said nobody ever.

As AI-based systems become more pervasive it's important to remember that these systems are for everyone and should benefit everyone. Sadly, this isn't always the case. Our very data is often biased heavily against different ages, genders, and ethnicities in ways that can be hard to detect.

Additionally, AI systems themselves can often be a new source of data from chat transcripts, user content, and more. How this data is treated matters greatly because even anonymized data may not be as anonymized as we think it is.

Finally, even when AI systems are working properly, they may be used in ways we don't intend or approve of, including ways that might cause real harm to others.

In this talk we'll talk about various ethical, privacy, safety, and fairness concerns in AI systems (generative and otherwise) and how to detect and remediate these issues so our AI systems can truly be for everyone.