The AI Partnership Project is the name Europe’s ETSI, Japan’s Telecommunication Technology Committee and South Korea’s Telecommunications Technology Association have given the global, industry-led AI specifications body they proposed on November 2, 2026. The three bodies signed a Letter of Intent at the Global ICT Standards Conference in Seoul, themed AI for All, to explore creating it together, and the proposal is explicitly modelled on 3GPP: the same organizational architecture, the same governance logic and the same theory of how to produce globally coherent technical specifications. The model is not accidental. In December 1998, representatives from seven regional telecommunications standards bodies met in Copenhagen to stop arguing and start collaborating. The result was the Third Generation Partnership Project, universally known as 3GPP, which eventually produced the technical specifications behind every smartphone on the planet. On November 2, 2026, three of those original bodies tried to repeat the trick for artificial intelligence. A Letter of Intent is a non-binding document that records shared intent to talk. It is not a standard, not a membership list and not a launched program. No benchmarks, funding figures or delivery dates were disclosed, and the AI Partnership Project’s scope remains undefined. ETSI Board Chair Markus Mueck, an Intel engineering director whose research focuses on 6G and AI integration, said the initiative can complement the existing AI ecosystem by providing a platform for industry, academia and research to align priorities. ETSI Director General Jan Ellsberger called the signing an important milestone in strengthening international cooperation on AI standardisation and invited partners worldwide to join. Bettina Funk, Chair of the ETSI General Assembly, said exploring a new partnership with Japan and South Korea offers ETSI members a route to practical standards.
Why the AI Partnership Project borrows the 3GPP model
The commercial opportunity is straightforward. Companies selling AI systems across borders must reconcile the EU AI Act, US state-by-state legislation in Colorado, Texas, California and New York, China’s algorithm governance rules, and national approaches in Japan and South Korea. A technically shared set of specifications could reduce duplicated testing, incompatible documentation and conflicting interpretations of safety and interoperability. It could also give governments a common reference point while leaving each jurisdiction responsible for adoption and enforcement.
3GPP supplies the template because it separates technical agreement from national adoption. Member companies draft and approve technical specifications through consensus groups. Regional Organizational Partners then transpose those specifications into their own systems, producing European Standards through ETSI, Japanese standards through TTC and Korean standards through TTA. The arrangement allows engineering work to proceed through the participation of companies and research organisations, while the regional bodies provide the formal path into national standards.
The AI Partnership Project would explore whether that architecture can work for technologies that change much faster than mobile networks. Telecommunications has historically relied on release cycles and stable deployment road maps. AI models, testing methods and software interfaces can change within weeks. Any proposed release mechanism would therefore need to preserve coherence without making implementation too slow.
There is also a question of what counts as a technical specification. ISO/IEC JTC 1/SC 42, the international AI standards body created in 2017, has concentrated largely on terminology and high-level frameworks. ETSI’s TC SAI has pursued more implementation-focused work, including EN 304 223, a European Standard establishing baseline cybersecurity requirements for AI models and systems. The Letter of Intent does not say whether AIPP would cover model documentation, safety testing, data provenance, interoperability protocols, cybersecurity or a broader portfolio.
The compliance gap creates urgency
The timing matters because companies cannot easily wait for every standards regime to mature. As of August 2026, when high-risk obligations under the EU AI Act took effect, the EU harmonised standards programme had delivered no standards with Official Journal references. Work by CEN/CENELEC under Standardisation Request M/606 has also slipped repeatedly, leaving businesses to prepare technical documentation, quality-management systems and conformity assessments without the full presumption of conformity that harmonised standards can provide.
That does not mean the European framework has failed. It means implementation guidance is arriving later than operators and product teams may have hoped. A body capable of producing clear testing protocols, system documentation requirements and evaluation methods could fill part of that operational gap. Its work would have value only if regulators and procurement authorities later referenced or incorporated the results, however. No government or standards organisation has committed to doing so at this stage.
South Korea has also been restructuring its domestic capabilities around the technology. TTA created four institutes in January 2026 covering 6G-AI convergence, on-device AI testing, full-stack AI verification, reliability certification for agentic systems, and AI applications across industries. These are concrete signs that the country sees standardisation as part of industrial policy and export competitiveness rather than a purely technical exercise.
Japan has placed greater emphasis on human-centric AI governance in international forums. That position is compatible with the language used by ETSI leaders, who repeatedly cited safety, sustainability, trustworthiness and interoperability. It may also help distinguish AIPP from an effort focused only on security or only on model performance. Yet broad principles do not automatically produce compatible technical documents, and the practical challenge will be deciding which requirements should be identical across markets and which must remain local.
Missing markets limit immediate reach
There is an important limitation to the proposal’s claimed global reach. ETSI, TTC and TTA are all Organizational Partners within 3GPP. The project’s five founding partners in 1998 were ARIB and TTC from Japan, ETSI from Europe, TTA from Korea and T1, later renamed ATIS, from the United States. China’s CCSA was not among the founders; it joined in 1999, and that expansion helped turn a regional partnership into a global one.
The United States and China are not signatories to the November 2 Letter of Intent. ATIS is not a signatory. South Korean conference coverage nevertheless describes ATIS as participating in wider discussions about AIPP, which is a different status from joining the signed instrument. The same distinction applies to China’s absence from the Letter of Intent, even though CCSA appears in the historical 3GPP founding record.
Without comparable standards organisations from the two largest AI markets, AIPP could still produce meaningful alignment across Europe, Japan and South Korea. Those regions contain major manufacturers, service providers, regulators and research institutions. They also have distinct priorities, including Europe’s risk-based legal framework, South Korea’s emphasis on technology competitiveness and Japan’s human-centered policy language. That combination gives the exploratory project a credible base. It does not give it the same reach that 3GPP achieved once participation expanded beyond its founding regions.
The venue reinforces the proposal’s connection to South Korean industrial policy. The Global ICT Standards Conference was held in Seoul from November 2 to 4 and co-hosted by the Ministry of Science and ICT with the Korean Intellectual Property Office. Coverage said the event attracted about 1,000 experts and would include 27 citations and certificates for contributions to AI and ICT standards. Sessions covered AI, quantum communications, satellite communications, physical AI, intellectual property and the development of standards-essential patents. KAIST professor Young-jae Jang was scheduled to speak on autonomous manufacturing and physical AI, while AIPP discussions formed part of the programme.
A Letter of Intent leaves the difficult questions open
No scope, budget, membership list or timetable has been agreed for the AI Partnership Project. The signed document establishes that the three organisations will explore options, not that a permanent body exists. There are also no disclosed benchmarks against which technical performance or governance quality could be measured, and no budget or funding plan for research, meetings, testing or secretariat operations.
Several design choices will determine whether the idea moves beyond discussion. Participants would need to decide which organizations can become members, whether voting rights would favour large companies or reflect wider stakeholder input, how research and civil-society expertise would enter the process, and how disagreements would be resolved. They would also need to define the relationship with ISO/IEC JTC 1/SC 42, CEN/CENELEC and established domestic bodies. A partnership may complement those institutions, but duplication could divide technical effort and confuse companies looking for one authoritative route.
Transposition will be equally important. In telecommunications, each Organizational Partner can adopt a shared specification through an established national process. AI rules do not yet have a comparably uniform structure. The EU AI Act relies on harmonised standards for parts of its compliance model, while national and regional systems outside Europe may use different terminology, testing regimes and legal presumptions. Any AIPP specification would need a clear route into those systems if companies are to treat it as more than a voluntary best practice.
Ellsberger invited partners worldwide to join, and the public interest in participation is clear. The challenge is converting that interest into a credible founding coalition without making the governance process so slow that practical standards arrive after the products they are meant to guide. South Korea’s conference emphasis on de facto standards reflects the same tension. Corporate specifications can move quickly because their users shape them, but they may also privilege the largest vendors unless competition and transparency are built into the process.
Ursula von der Leyen made the international case for cooperation at the Tallinn Digital Summit, arguing that European standards become global through engagement with trusted partners. ETSI’s Letter of Intent is a modest step in that direction, involving Europe, Japan and South Korea rather than a truly worldwide coalition. Its value lies in testing whether the 3GPP method can be adapted to a faster and more politically fragmented technology. The AI Partnership Project could eventually provide common technical building blocks for trustworthy and interoperable AI, but at present it remains an exploratory proposal without agreed scope, budget, membership list or timetable.