Callers would have to say within five seconds that a machine is speaking.

Businesses and public bodies that use software to make telephone calls would have to tell the person answering, within five seconds, that they are speaking to a machine, under draft rules published by the regional communications regulator on Tuesday. The proposals cover synthetic voices produced by artificial intelligence as well as older recorded-message systems, and would apply to both sales calls and service calls such as appointment reminders.

The regulator said complaints about automated calls had risen from about 2,300 two years ago to more than 9,000 in the past twelve months. A growing share involved voices so natural that callers did not realise they were not speaking to a person until well into the conversation, and in some cases not at all.

What the draft says

Under the proposals, any call initiated or conducted by software would need to open with a short spoken statement, such as “This is an automated call”, before any other content. The statement would have to be repeated if the caller asks whether they are speaking to a human. Organisations would also have to offer a way to reach a person, either by pressing a key or by saying so, and the option would have to be announced in the first thirty seconds.

  • Disclosure within five seconds of the call connecting.
  • An honest answer whenever the person asks if the voice is human.
  • A clear route to a member of staff during working hours.
  • Records of automated calls kept for twelve months and handed over on request.
  • Fines of up to 4 per cent of regional turnover for repeated breaches.

Calls made by emergency services and some public health alerts would be exempt, as would systems that simply connect a human caller, such as call-routing menus that answer incoming calls rather than make outgoing ones.

A divided industry

The response from companies that build voice systems was mixed. Several larger firms said they already disclose automation at the start of calls and welcomed a common standard. Smaller developers, including a number based in the valley’s technology park, warned that a strict five-second window would be hard to meet when calls connect to voicemail or when the line is poor.

“Nobody serious in this industry wants to fool people. But the rules have to cope with the real world, where half the calls land on an answering machine and the first two seconds are silence,” said Rafael Ondine, chief executive of Tessel Voice, a twenty-person firm in the technology park.

Consumer groups argued the opposite: that five seconds was generous. The Calder Consumer Forum said its volunteers had recorded calls in which a synthetic voice made small talk, apologised for background noise and laughed at its own jokes, behaviour it said was designed to suggest a human was on the line.

Who would enforce it

The regulator would gain powers to request call logs and audio samples from companies suspected of breaking the rules. Telephone networks would be required to block numbers linked to repeated breaches. Officials acknowledged that the hardest cases would involve callers based outside the region, and said they were in talks with counterparts elsewhere about sharing information.

Celia Marchbank, the regulator’s director of consumer protection, said the aim was to restore a basic level of trust in the telephone. “When you answer a call, you should know whether there is a person at the other end,” she said at the launch. “That is not a technical question. It is a question of respect.”

The consultation runs for twelve weeks. If adopted, the rules would take effect next autumn, with a six-month grace period for organisations to update their systems. The Assembly’s digital affairs committee has asked to hear evidence from both the regulator and industry representatives in January, and members from several parties have indicated they would like the requirements extended to text messages and chat services.