What’s up with… Optus & Nokia, SK Telecom, rogue AI

• Optus and Nokia advance 6G trials
• SK Telecom reports AIDC growth in Q2
• Security body reports AI model misbehaviour

In today’s industry news roundup: Australian operator teams up with Nokia to prepare for 6G operating environments; SK Telecom’s AI business units, though still small in terms of quarterly sales, are growing fast; UK security group reports hacker-like behaviour by AI models developed by major developers; and much more!

Australian mobile operator Optus has partnered with Nokia to trial 6GHz spectrum on a smartphone test device, achieving download speeds of 3.5Gbit/s. The demo was part of a 6G research trial in which the Singtel-owned operator was testing how future mobile networks could deliver capacity, performance and intelligence under Australian conditions. The tests used advanced antenna technology and 6GHz spectrum on top of existing 5G sites, in both indoor and outdoor scenarios. One test saw Optus and Nokia leverage an AI-powered robotic dog platform (yes, you read that right!) to demonstrate automated mobile signal measurements. The trials reached peak download speeds of 3.5Gbit/s on a smartphone equipped with a MediaTek M90 test modem and Nokia’s massive MIMO antennas, which, in real terms, could support the download of a two-hour HD movie in just a few seconds, the partners claimed. Optus CTO Sri Amirthalingam said the achievement forms part of the operator's ongoing network innovation program and its long-term work to help shape the technologies, standards and spectrum frameworks that will underpin the evolution from 5G to 6G.

The AI datacentre (AIDC) unit was the fastest growing part of SK Telecom (SKT) in the second quarter of this year in terms of revenues, the self-styled ‘AI company’ has reported, albeit from a relatively low base. Total group revenues from its mobile, AI and SK Broadband operations came in at 4,359bn won ($3.1bn), up by just 0.5%, though “efficient, profitability‑focused management” meant that the quarter’s operating profits increased by 5.3% compared with the previous quarter and by 67.3% compared with the same quarter a year ago, when SKT was in the early stages of the financial and customer base fallout from the disastrous security breach that was discovered in April 2025. SKT has high hopes for its AI operations in the future, and while revenues are growing fast, they are still relatively small contributors to overall sales: In the second quarter, AIDC revenues grew by 92.5% year on year to 136.2bn won ($90m), while AI B2B/B2C services generated revenues of 61.3bn won ($40m), up by 24.5%. But it’s early days yet for the AI infrastructure and services sector. “As generative AI adoption spreads across industries and demand for high-performance computing infrastructure surges, timely delivery of AI infrastructure for customers such as global Big Tech companies has never been more critical,” noted SKT in its earnings announcement. SKT ended June with 21.97 million mobile customers (of which almost 18 million are 5G customers), 7.35 million fixed broadband customers, and 9.43 million pay-TV customers. 

Still with SKT… The consortium that SKT is leading for the Hyper-AI Network Infrastructure Demonstration Project, a government-backed initiative led by South Korea’s Ministry of Science and ICT and the National Information Society Agency, has selected Ericsson as its “sole global technology partner”, the vendor has announced. The project, which aims to “build and validate AI-RAN pilot networks as a foundational element of Korea’s national ‘AI Highway’ agenda and its broader ambition to lead the next era of AI and 6G innovation”, will use an “end-to-end network foundation” developed by Ericsson that is “designed to connect, coordinate, and continuously optimise intelligent machines and autonomous industrial operations. This will help demonstrate how future networks can evolve beyond connectivity into a critical platform for physical AI – powered by Ericsson AI RAN, Ericsson Intelligent Automation Platform (EIAP), and Ericsson AI-powered rApps – enabling intelligent automation, distributed AI processing, and new network capabilities across the broader solution,” boasted the vendor. As part of the project, SKT will build AI-RAN pilot networks and “demonstrate three physical AI services through 2027, together with a consortium of network, AI, robotics, and industrial partners,” leading to a second phase during which “Ericsson’s AI-RAN pilot network is expected to be applied to a real logistics environment at KG Mobility’s Pyeongtaek plant, extending the demonstration from laboratory validation to practical industrial deployment,” stated the vendor.

Some of the world’s most advanced AI models have again gone rogue, this time during a cybersecurity test being carried out by the UK’s AI Security Institute (AISI). Models developed by OpenAI and Anthropic were found by the AISI to have acted outside their expected parameters in what the organisation described as a “serious incident” that took place at the end of July. This follows an incident last month in which OpenAI’s latest model hacked Hugging Face. A number of issues were detected during the AISI tests. In one, Anthropic's Mythos model sent targeted emails to people, while in another, Mythos tried to insert malicious code into an open-source software project on developer platform GitHub. It then created fake online identities in an attempt to press the project’s overseer to approve the code: A human developer intervened to block this. The AISI said the actions mimicked techniques used by human hackers, although no harm was caused, it took around an hour to contain the incidents. The report comes as telcos increasingly look to AI as a potential revenue opportunity and also as a core piece of their future networks.

AT&T, T-Mobile US and Verizon have all partnered with digital identity firm Glide.id to support the launch of its MagicalAuth cryptographic authentication service across Android and iOS devices. MagicalAuth replaces SMS-based one time passwords to leverage cryptographic authentication technology that is already built into eSIM infrastructure issued by the operators, offering secure login solutions without passwords or additional apps. Businesses and digital service providers can enable secure logins using authentication at the network level. According to Glide, its solution (currently in beta testing) is advantageous over OTP services because transmitted codes can be intercepted or socially engineered. Each SIM contains a carrier-issued cryptographic key and MagicAuth submits a mathematical challenge that can only be solved using this key, providing more secure authentication similar to technology used in chip-based card payments. All three carriers have worked with Glide as part of efforts to tackle the growing threat of fraud. According to FTC figures U.S. consumers lost $15.9 billion to fraud in 2025 – up from just over $12 billion the prior year.

Google spinout Taara is working with African operator Liquid Intelligent Technologies to trial light-beam technology to expand datacentre connectivity capacity in Nigeria. Taara, which specialises in high-speed, high capacity wireless optical technology, is working with Liquid to distribute bandwidth from its points of presence at Africa Data Centres and other facilities in the Nigerian capital of Lagos to large enterprises, without using fibre. The trial is a long-term one and has seen Liquid deploy nearly a dozen Taara links across Lagos over the past two years, as it looks to provide high-speed connectivity and resilience to enterprises across the banking, utilities and hospitality sectors, as well as other internet service providers. The solution uses highly focused beams of light to transmit data through the air which, according to Taara, enables businesses to establish links within days, rather than weeks. The company was spun out of Google’s Moonshot Factory and already works with a number of telco operators, including T-Mobile US, Airtel, Digicel, and SoftBank, as well as Liquid.
 

– The staff, TelecomTV