ML systems corruption and security

If your ML system lies on external data you gather and process. Be careful to not let your competitors pollute these sources with dump data that will make your system go crazy and giving bad predictions. If your sources of data are user-generated this will be easier for them to register to these sources as users and polite the data. Data pollution will be an issue for the coming years. Spoofing is already used in the financial markets by sending many orders and canceling them right after. Reg NMS brought some solutions to this problem the same will be true with internet products, services, and marketplaces. Be careful about your ML system because your competitor can corrupt it.