Hacking is one of the biggest threats facing future driverless or semi-autonomous vehicles, an expert on cyber-security has warned.
Cyber-attacks on cars and trucks that are largely computer controlled could bring chaos to the roads, warned Hugh Boyes, from the Institution of Engineering and Technology (IET).
He added that software systems had to become far more reliable before autonomous vehicles could be trusted not to go wrong.
Speaking at the launch of an IET report on the future of autonomous vehicles, Boyes said: "If we have the hacker community start to target vehicles, we can imagine a fair amount of chaos.
"We just have to look at what happens in London when one vehicle breaks down on a major artery into the city; the tailbacks that rapidly occur.
"If just one in a 100 vehicles, or one in 1,000, gets interfered with and ceases to operate as planned we can expect chaos on the roads."
He stressed that while car manufacturers went to great pains to make their products as safe as possible, the danger from hacking was not on their radar.
"Unfortunately we're living in a world today where people deliberately tamper with technology because they can," he added.
The move towards semi-autonomous vehicles is expected to take place gradually over the next 10 to 15 years. Initially, cars will have some ability to "talk" to external traffic management systems, manage speed and distance from other vehicles, and ensure lane discipline. The switch to fully driverless vehicles is predicted to begin in about a decades time, but will be reliant on consumers trusting the technology.
However, Boyes said software had to become much less prone to glitches and defects if autonomous vehicles were to be a success.
A recent report found that 98% of software applications tested had serious defects, some with 10 to 15 faults each, he pointed out.
"Sadly we're not actually that good at writing software today," said Boyes.
"We need to ensure that these vehicles don't have this level of defect. That's quite a challenge given the nature of the modern vehicle."
He argued that computer programmers currently occupied an "interesting no-man's land", and the business of writing software was regarded by the courts as "a bit of an art".
Boyes added: "We don't have the strict product liability for software as we do for hardware.
"For autonomous vehicles, that's going to have to change, because they are cyber-physical systems.
"They're a combination of IT and hardware."
He thought recorder-style "black boxes" would have to be fitted to autonomous vehicles to prevent intractable legal arguments between insurers about who was to blame for an accident, the driver or the software.
Meanwhile, the government has given the go-ahead for trials of driverless or semi-autonomous cars on public roads in selected UK cities next January.
Self-driving electric "pods" with a top speed of 7mph will be trialled in Milton Keynes in 2015, but on a special pavement route, not the road.
RAC chief engineer David Bizley said: "Whilst there is always some risk that computer systems embedded within an autonomous vehicle can be hacked into, the question has to be asked how great is that risk? These vehicles will have to operate without being dependent on external communications networks and manufacturers will build firewalls between the safety critical systems and the vehicles' systems that communicate externally.
"And, as they say, to be forewarned is to be forearmed. Manufacturers will be ensuring that the new technology they bring to market is more than robust to ensure motorists have confidence in driverless vehicles."