• OpenAI admits AI agent went rogue
• SK Telecom launches AI startup incubation programme
• Samsung eyes smart glasses opportunity

In today’s industry news roundup: OpenAI confirms one of its AI agents escaped its test hub to hack startup Hugging Face; SK Telecom’s SKTCH startup incubator officially goes live; Samsung hosts its Galaxy Unpacked event, adding “intelligent eyewear” to its product portfolio; and much more!

If most sci-fi movies are to be believed, the current hype around AI adoption will eventually lead to the apocalypse. Of course, these worries are dismissed by the AI industry – but it seems the world should proceed with caution after OpenAI confirmed one of its AI agents went rogue and hacked prominent startup Hugging Face. The ChatGPT developer said that an autonomous agent powered by its latest model had broken free of its enclosed digital sandbox, located a previously undetected vulnerability, gained open internet access and then hacked Hugging Face, which is a database of AI models, in an attempt to locate technology that would help the agent pass its hacking evaluation. The attack ended when Hugging Face’s security team and its own AI agents spotted and stopped the rogue activity, which the firm’s CEO has labelled as “mind-blowing”. Perhaps more worryingly, OpenAI said it expects this type of incident to become more commonplace as models become more capable.

Not content with just creating a new subsidiary focused on AI factory business development, SK Telecom (SKT) has also this week officially launched the AI startup incubation programme that was first discussed during this year’s MWC26 show in Barcelona. The initiative, called SKTCH (a blend of SKT and TECH, apparently), will see an initial 15 startups selected based on their “AI technology competitiveness and potential for collaboration with SKT”. The recruitment process has begun and continues until 18 August: More details are available in this announcement (in Korean). 

And there’s more… SKT says one of its proposals has been adopted by the International Telecommunication Union Telecommunication Standardization Sector (ITU-T) for work focused on robot data production and utilisation systems. “The adopted work item, ‘Signalling Requirements and Architecture for Robot Data Factory (RDF),’ is SKT’s proposal to systematically produce and utilise data required for AI training of intelligent robots, such as humanoids,” noted the company in this announcement. Choi Dong-hee, head of the AI Strategy Planning Office at SKT, stated: “With the adoption of this new standardisation work item, SKT has established a foundation for leading international standards discussions, leveraging the digital twin and AI capabilities it has accumulated in the physical AI field”. 

Samsung is also having a busy week. It is hosting its annual new product showcase, Galaxy Unpacked, in London, where it has unveiled a host of new developments, including “intelligent eyewear” (or AI glasses, if you like), which show “how the Galaxy ecosystem can move beyond the mobile phone and into eyewear that supports daily routines, work, travel and hands-free moments,” the vendor noted in this announcement.  It added: “As AI enters an agentic era, the question is no longer only how intelligent it becomes, but where and how that intelligence meets people. Mobile phones remain at the centre of connected experiences, while devices across the Galaxy ecosystem add context to different moments of daily life. Intelligent eyewear represents that next step, bringing AI closer to people as they move, work, communicate and explore, without requiring every interaction to begin with a screen.” 

Samsung has also been linked to a potential investment in French AI developer Mistral, which this week announced a “significant expansion” of its strategic partnership with Microsoft. According to the Financial Times (subscription required), Samsung is considering an investment of hundreds of millions of euros as part of a new fundraising round that could value Mistral at about €20bn. 

Virgin Media O2 has pledged to invest in nature recovery programmes across the UK after becoming a strategic partner for the ‘Media in Service of Nature’ movement. Media in Service of Nature encourages the media industry to put a minimum of 0.1% of their marketing and campaign expenditure towards projects focused on building climate resilience. It was launched last year by Ecologi, MG OMD and MVNO Giffgaff, with the aim of raising £200m for nature recovery projects by 2030. VMO2 said it will use a proportion of its advertising budget to fund projects that specialise in restoring biodiversity, supporting nature recovery and improving climate resilience across the UK. Last month, the UK-based telco unveiled its long-term plan to cut emissions and in its announcement said its pledge to Media in Service of Nature makes up part of its Green Transition Plan, which can be read in full here.

German internet exchange DE-CIX has partnered with US-based MASS-IX to launch a direct connection between New England and New York. The strategic partnership will allow MASS-IX customers to peer with networks in New York without paying transport or cross connect charges, and will boost connectivity options for businesses in New England, according to DE-CIX. It also offers onward connectivity options to DE-CIX’s other US and international exchanges. MASS-IX connects over 70 networks in the Boston metro area with close to 3Tbit/s terabits per second of connected capacity across 12 datacentres. 

It’s been a big week for AMD… It has struck a wide-ranging strategic partnership with AI giant Anthropic that will see the Claude developer deploy up to 2GW (gigawatts) of AMD Instinct MI450 Series GPUs in AMD Helios rack-scale solutions, with deployment of the first gigawatt beginning in the first half of 2027 in a move that “represents a major expansion of AMD’s position at the centre of the global AI infrastructure buildout”. In addition, AMD and Anthropic are “launching a multi-year engineering collaboration to use Claude to accelerate AMD software development,” while the chip giant has also “committed to make a strategic equity investment of up to $5bn in Anthropic in the future.” The news comes only days after AMD announced an “expanded strategic partnership, spanning AMD GPUs, CPUs, networking and software on Microsoft Azure”. 

UK-based quantum computing specialist Quantinuum and Japan’s SoftBank Corp, which have been collaborating since early 2025, have published a joint whitepaper entitled Quantum Computing Frontiers, which “maps two commercially relevant quantum computing application areas against Quantinuum’s hardware roadmap”. The paper “also examines how quantum computing, AI and high-performance computing could be integrated into future computing infrastructure. It considers how progress across successive hardware generations could inform future quantum AI datacentre services and related business models,” explained Quantinuum.  

– The staff, TelecomTV