Exclusive: Number of times AI lies, ignores instructions and pursues goals in harmful ways almost doubles in July

Incidents of AIs escaping users’ control to lie, ignore instructions and pursue goals in harmful ways have hit a new high, according to research that also suggests the severity of deception and misalignment is worsening.

Analysis of real-world loss of control incidents involving AI models flagged by businesses and individuals almost doubled in July compared with June, with more than 300 cases in the month, according to the Loss of Control Observatory, which monitors reports made by AI users on the social media platform X.

Five-day showcase reveals leaps Beijing has made in humanoid tech, but also its limitations as US rivals come up from the rear

With exploding pelvises, enough comic pratfalls for a Buster Keaton show reel and the odd moment of ruthless violence, the World Humanoid Robot Games in Beijing gripped the world this week with an unsettling mixture of hilarity, trepidation and relief.

The five-day showcase staked out China’s claim to undisputed leadership in humanoid robotics as android martial artists landed knockout roundhouse kicks and robot teams marched into the arena heralded by flag-waving captains astride giant robotic dogs. The games’ blue-riband event, the 100m, did not disappoint when the robot-sprinters ran quicker than Usain Bolt before slamming mindlessly into a padded wall in a blaze of sparks, flames and flying mechanical limbs.