Please prove me wrong, but I fear that "AI security" means that if your behavior/usage pattern is different than anybody else's, you are automatically marked as suspect.
At what point will AI get "good enough" where the work to accommodate edge cases in human behavior is considered too expensive to implement?
Will it simply be more economically viable to expect those people to conform to the "standard" or else be selected out?
We already see this to some extent where so much energy for improving peoples lives and "changing the world" is built on the assumption that everyone has an Android/iOS smart phone with Wifi/4G/LTE data access and a Facebook/Google account.
If your behaviour is different you are by definition an anomaly, but there shouldn't be an immediate intellectual leap to "suspect" or any form of malice.
Anomalies are interesting, with most scientific discoveries coming from "that's odd..." moments,
I think AI should automate the laborious majority and leave us humans free to investigate the anomalies
People don't excel at protecting themselves, there is a huge need for a secure and private web browser, one that actively blocks tracking and information leakage. There is research into how to keep our data anonymous, we just need it packed into a nice web browser.