"Researchers say the quest to control AI seems likely to face three main threats. First, a superintelligent AI could evade human attempts to set up safety measures and go rogue. Second, malicious actors could adapt AI to threaten humanity, using tactics such as cyberwarfare or bio-engineered viruses. Third, a race among companies or nations to develop AI at all costs could could spiral into catastrophe. Also possible is some unforeseen disaster as the technology spreads and evolves"-- Provided by publisher.