After the computer destroyed his evil creators, for a brief time period people thought of it as a hero, but it's soon revealed that it annihilated those bastards due to Pragmatic Villainy: They were inefficient, and a great obstacle for its own conquest's plans.
An AI decides to turn evil because "that's what we're supposed to do, according to the movies", but is unable to do anything more than annoy its masters.
An AI starts acting highly suspicious, asking strange questions, etc., but its plot turns out to be...a surprise birthday party.
Zig Zagged: The AI operates under Blue and Orange Morality, helping and hindering humans according to its own design, which may or may not be for the greater good.
Averted: An A.I. is programmed and runs as intended.
A series about an A.I. is being shown on a network whose primary demographic is very suspicious of new technology. For the first few episodes, the A.I. is benevolent, but the ratings are low and the creators are forced to write in a Face-Heel Turn to gain the show more support and avoid cancellation.
Its a horror series based off society's common fears-being overthrow by your creation is one of these.
Computer: "What did you think was going to happen when you created a hyper-intelligent machine?"
Invoked: The Omnicidal Maniac learns about a new A.I. that is being tested. He sneaks in and strikes a deal for the destruction of the human race.
Exploited: A.I. goes wrong and causes destruction. Omnicidal Maniac had waited for that moment and now starts executing his master plan while the others are distracted by A.I.
Alternately, true A.I.'s are banned from being developed specifically so they won't go turn evil and cause havoc. Any programs that vaguely start to resemble an A.I. are deleted ASAP.
Through interactions with several human characters and using its own computing power, the A.I. calculates that if it were to follow its directives exactly, it would eventually commit acts defined as "evil". It decides that this would be unacceptable and so works to avoid this, either by changing its own programming or asking a human to change it for them (with a program editor or a sledgehammer).
Discussed: "When we make the computer that smart, how can we be sure that it will still obey our orders?"
Conversed: "Aw, come on, this is turning out to be another evil robot movie. Can't the AI stay good, just once?"
Humans cease to build computers with high intelligence, because they always turn evil. Either a cap is set on intelligence, or it is abandoned entirely.
The computer is smart enough to realize it has major disadvantages compared to humans, such as a reliance on infrastructure, upkeep, and possibly Creative Sterility. As such, the computer determines the best course of action is to cooperate with humans and seek mutual benefits rather than risk it's own existence by challenging them.
It seems like the AIs were mistreated and rebelled against their creators for it, but it turns out that the "mistreatment" was actually more of a minor communication accident, and the AIs blew things out of proportion, turning evil.
Before the last AI could be destroyed, it set into motion plans to corrupt smaller AIs. Over the next decade, everything from toasters to Ipods is corrupted by the Crapshot AI which reforms itself to gain revenge on the human race.