"Machine learning security for software developers"

AI is riding the hype cycle nowadays. It beats us at video games, it mimics artists to an uncanny degree -- and most importantly, it is also trying to write, break, and secure source code. Hackers everywhere are probing systems, asking "What can possibly go wrong?" As it turns out, a lot of things: with a little interference, self-driving cars can ignore stop signs, security cameras will fail to notice a burglar, chatbots will inadvertently leak credit card information, medical systems will prescribe lethal doses for the patient, and Twitter bots will become toxic and hateful.

We'll be exploring this AI threat landscape, especially some less-discussed parts -- the software security of ML systems, since they run on code like everything else. Just how much do you trust your model zoo? How many vulnerabilities are in your favorite ML framework? Will ChatGPT usher in a new era of robotic script kiddies? And does your favorite AI know about Little Bobby Tables?