The government is preparing legislation intended to make it impossible for anyone under 18 to take, view or send nude images on smartphones and tablets.
It has not published the accuracy standard that the technology for detecting images would have to meet.
Nor has it set out how many legitimate images could be wrongly classified before a system designed to block rather than warn would be considered unacceptable.
The proposed law would require nudity detection to be switched on by default for children and apply across device features including cameras and apps made by third parties. Adults would be able to disable the restrictions only after proving they are over 18 through what government describes as robust age assurance.
Some of the underlying technology already exists.
Apple's Communication Safety analyses photographs and videos on the device and can blur content it identifies as containing nudity. It is switched on by default for users under 18 on supported devices. For children under 13, a parent or guardian using a Screen Time passcode is required before detected material can be viewed. Teenagers can still proceed after warnings. Apple says the image analysis takes place locally and it does not receive an indication that nudity was detected unless the user actively reports content.
Google's Sensitive Content Warnings work on a similar principle in Google Messages. SafetyCore classifies material on the device, blurs images that may contain nudity and warns users when they attempt to send, receive or forward them. Google says neither the classified content nor the classification result is sent to its servers.
Neither system currently provides the blanket prohibition the government is proposing.
The government said in June that Apple's protections did not cover the camera, the wider range of apps made by third parties, messaging services or search functions comprehensively enough to meet its objective. It gave Apple and Google three months to strengthen their systems voluntarily and said legislation would follow if they did not.
On 9 September ministers said the companies had made progress but had not gone far enough, and legislation would now be prepared. The government also said it could reconsider the need for legislation if companies implement sufficient protections themselves.
The move from a warning system to a prohibition makes accuracy more important.
Conservative MP Wendy Morton asked the Home Office in June what technical standards would apply and whether government would publish an independent assessment of accuracy and false positive rates.
Safeguarding minister Natalie Fleet said discussions with industry would include "consideration of accuracy thresholds". The answer did not specify a threshold or commit government to publishing the independent assessment of false positives Morton had requested.
Government also has not published a detailed classification standard explaining how the proposed system should deal with nudity that is not sexual or other material that an automated classifier could identify incorrectly.
That could include medical photographs, breastfeeding images, art or family photographs. Existing Apple and Google systems can place a warning or blur over material they classify as sensitive; the government's stated objective is that a user under 18 should be unable to proceed when nudity is detected.
Age assurance is already more developed than the original announcement might suggest.
Apple introduced UK age confirmation measures in 2026. Adults can confirm they are over 18 using methods including an eligible credit card or identity documents. Apple says Communication Safety and web content restrictions are switched on automatically for children, teenagers and adults who have not confirmed they are over 18.
Apple also allows an age range rather than a birth date to be shared with participating apps. Its developer system can use information confirmed by payment methods, government identification or other methods while limiting the personal data passed to the app.
The government has not said that Apple's system will become the national standard, or specified how equivalent verification would work across Android devices and applications made by third parties.
Age assurance itself is not a solved technical problem.
A Department for Science, Innovation and Technology study published in March found that facial age estimation systems have improved but still struggle around narrow age boundaries such as 13 and 18 and can show differences in performance by gender, skin tone and ethnicity. One vendor tested achieved an average error of around 1.2 to 1.3 years, which the study found still meant a child of 12 could be mistaken for 14, or one of 16 for 18.
The government's wider online safety plans acknowledge those problems. Its June consultation response says age checks need to become more accurate and that Ofcom has been asked to set clearer expectations for highly effective age assurance.
There is also no published technical answer yet to every route by which the proposed nudity restrictions could be bypassed.
Morton asked specifically about VPNs, sideloaded apps, encrypted messaging, cloud storage, web browsers and older devices.
The Home Office said operating system controls should provide the broadest coverage and that it would continue examining potential circumvention. It said the policy would cover new and existing supported devices through operating system updates, but did not explain how each of the routes Morton identified would be prevented.
That qualification is particularly relevant to older phones. The government referred to "maximum coverage" of devices already in circulation rather than universal coverage, while limiting its answer to devices capable of receiving supported operating system updates.
The proposal does not inherently require encrypted messaging to be broken.
Images can be classified locally before they are encrypted for transmission or after they are decrypted for display. Apple's Communication Safety already performs local analysis while maintaining end to end encryption.
The policy would instead give software at device or application level greater authority over whether an image can be captured, displayed or transmitted.
Government is considering obligations on apps as well as operating system providers, meaning the eventual legislation could require protection at more than one layer of a device. The September announcement does not yet specify how responsibility would be divided where an operating system, app and online service produce different classifications.
The child safety problem behind the proposal is well documented.
The Internet Watch Foundation says it took action on 311,610 reports containing child sexual abuse material in 2025, a 7 per cent increase on the previous year. Self-generated imagery disproportionately affects teenagers: of 63,044 teenagers aged 14 to 17 who appeared in 56,179 child sexual abuse images identified during the year, a third were self-generated, a term the IWF uses for imagery in which a child appears without an abuser physically present and which can result from grooming, coercion or extortion.
The IWF recorded 397 reports of sexually coerced extortion involving child sexual abuse imagery in 2025, up 127 per cent from the previous year. Boys aged 14 to 17 accounted for 98 per cent of those cases.
The IWF supports protections on the device itself and has argued that nudity detection could play a significant role in preventing images from being produced and circulated in the first place.
The government therefore has evidence for intervening before images spread online rather than relying entirely on platforms to remove them afterwards.
What remains unpublished is the technical benchmark against which that intervention would be judged.
In June, Morton separately asked what criteria would determine whether Apple and Google had met the government's demand to act within three months. The Home Office said success would be aligned with achieving "comprehensive nudity detection" across smartphone and tablet operating systems and that trends in self-generated child sexual abuse material would be monitored. It did not publish a numerical accuracy requirement or tolerance for false positives.
The September announcement retains the same absolute objective: children under 18 should be unable to take, view or send nude images.
It also acknowledges that the technical work is unfinished, saying the complexity of the issue means legislation must be developed carefully and promising introduction "as soon as possible".
No legislation has yet made it technically impossible for children under 18 to create or share nude images.
Apple and Google already demonstrate that private classification of nudity on the device is possible, and Apple has introduced UK age confirmation measures. What government has not yet published is the standard that turns those existing warning systems into the blanket prohibition ministers have promised: the required detection accuracy, acceptable rate of false positives, treatment of legitimate images, coverage of unsupported devices and the test for whether circumvention has been prevented.