On 4 June 2026, the Judicial Court of Caen found that Carrefour's online grocery website and mobile app were not sufficiently accessible to visually impaired users, which shows what smart technologies for social impact are judged against once the law is enforced. Taylor Wessing describes it as a landmark interim ruling, and it came a year after the European Accessibility Act (EAA) became enforceable across the EU on 28 June 2025.

Carrefour's own audit had shown 71.21% conformity with EN 301 549 (the European harmonized technical standard that defines accessibility requirements for ICT products and services), assessed through France's RGAA framework.

The court still found significant barriers, including missing or inadequate alternative text, weak colour contrast and incompatibility with assistive technologies and it required each criterion to be assessed individually. Carrefour faces a penalty of €500 a day if it is not fully compliant within six months.

Where automated testing stops

Regulators already use software to find such faults. Germany's market surveillance authority screens web services with technical testing tools, and AudioEye's scan of more than 166,000 pages found EU sites averaging 25% more issues per page than US sites.

AudioEye sells accessibility tools, so the figure is best read as indicative. Scanners can catch poor contrast and unlabelled links. Judging whether alternative text is useful or whether a screen reader user can finish a purchase, needs a person with assistive technology.

IEEE 7000-2021 -the first international standard that provides a systematic, repeatable process for embedding ethical values directly into system design and engineering- offers a method for this. It sets out a process for eliciting stakeholders' values and tracing them into system requirements from the concept stage, so values are part of the specification from the start.

The AI4People framework makes a related point for artificial intelligence, adding explicability to the four established principles of beneficence, non-maleficence, autonomy and justice. A decision that people can inspect criterion by criterion is one they can contest.

The cost of skipping this step is visible in the UK. The National Audit Office found that take-up of GOV.UK Verify was much lower than expected and its chief auditor said only half the people who tried to sign up could use it. The government announced it would stop funding Verify from March 2020.

WCAG 2.2 targets this kind of friction. The European Commission's AccessibleEU centre notes that its accessible authentication criteria bar login methods that rely on cognitive function tests, such as solving puzzles or transcribing text, and that it sets a minimum target size of 24 by 24 CSS pixels.

Blockchain as civic infrastructure, with caveats

Testi, Marconi and Pasher's review in Open Research Europe argues that blockchain can support transparency, trust and citizen engagement in smart governance. It favours consortium architectures with a Proof of Authority consensus algorithm, and it acknowledges challenges of scalability and security.

Estonia is the usual example, and a Belfer Center case study describes its KSI blockchain as a permissioned system for data integrity. The same study cautions that blockchains can strengthen institutional accountability without automatically decentralising power.

One practical point remains. Taylor Wessing reports that a revised EN 301 549, expected to reference WCAG 2.2, was due for publication in the third quarter of 2026. Organisations still testing against WCAG 2.1 should check whether it has appeared.

Sources and further reading

• Taylor Wessing, "The EAA and BFSG: recap and update" (30 June 2026)
• AudioEye 2026 Digital Accessibility Index coverage
• IEEE 7000-2021 launch announcement
• AI4People ethical framework summary (Atomium-EISMD)
• National Audit Office, "Investigation into Verify"
• European Commission AccessibleEU: WCAG 2.2 is officially a W3C Recommendation
• Testi, Marconi and Pasher, Open Research Europe
• Belfer Center: Estonia case study