Audio-first model
Indian startup Sarvam AI has launched Sarvam Audio, an audio-first language model designed to understand and transcribe real-world speech across India’s multilingual landscape. Built on the company’s 3-billion-parametre Sarvam 3B model, the system supports transcription in 22 Indian languages, including Hindi, Tamil, Telugu, Malayalam, Bengali and Indian English. Unlike global rivals such as ElevenLabs, which focus on expressive voice generation, Sarvam Audio prioritises speech understanding and transcription. The company claims the model outperforms GPT-4o-Transcribe and Gemini-3-Flash on the IndicVoices dataset, achieving lower word error rates across unnormalised, normalised and code-mixed transcription styles.
British Museum faces flak
The British Museum has removed social media posts after archaeologists criticised it for sharing AI-generated images. The posts, uploaded on January 27 to Instagram and Facebook, showed a digitally created young woman contemplating museum exhibits and were tagged to an AI model and marketing agency V8 Global. Critics questioned the ethics of heritage institutions using AI-generated visuals without transparency. In response, the museum said it regularly shares user-generated content but acknowledged the sensitivity around AI imagery. “We do not post AI-created images and, recognising the potential sensitivity, removed it,” a spokesperson said.
A hot spring that didn’t exist
An AI-generated travel recommendation led tourists to search for a fictional attraction in Tasmania. A blog on the Tasmania Tours website promoted ‘Weldborough Hot Springs’ as a serene natural retreat, prompting visitors to travel to northeast Tasmania only to find no such destination existed. The error damaged trust among travellers and raised concerns about unchecked AI-generated marketing content. Scott Hennessey, owner of Australian Tours and Cruises, which operates the site, admitted the company’s AI system “messed up completely”. The issue emerged after content creation was outsourced to a third-party provider.
L’Oreal’s tech hub
L’Oreal has announced plans to establish its first global AI-powered beauty technology innovation hub in Hyderabad. Revealed at the World Economic Forum, the facility will form part of the company’s Global Capability Centre strategy. L’Oreal expects to invest Rs 3,500 crore by 2030 and create around 2,000 specialised roles in AI, data science, engineering and technology. Hyderabad will become one of L’Oreal’s four regional technology hubs, responsible for integrating AI into digital platforms and customer solutions.
Age no bar for AI
In Shenzhen, China’s technology hub, retirees are increasingly turning to AI through evening classes designed for older learners. At an AI night school, elderly students learn to use tools such as Jimeng, a Midjourney-like image generator developed by ByteDance, to restore and animate old photographs. The programme has attracted strong interest, supported by more than 200 volunteer instructors. The initiative, which highlights an unexpected dimension of AI adoption in China, extending beyond startups and youth to include seniors.
Criminalising obscene images
Lawmakers in the US state of Georgia are considering legislation that would criminalise the use of AI to create obscene images of real people without consent. Senate Bill 398 proposes a new offence termed ‘virtual peeping’, targeting AI tools that digitally undress or manipulate images of individuals. Sponsor Senator Bo Hatchett described the technology as a growing ‘nightmare’ for victims. Under the bill, generating obscene AI images of adults would be a felony punishable by up to 10 years in prison and a $50,000 fine. Cases involving minors would carry harsher penalties, including up to 20 years’ imprisonment and fines reaching $100,000.
Content marketplace
Microsoft has introduced the Publisher Content Marketplace (PCM), a new platform aimed at licensing premium publisher content for AI training and grounding. The initiative offers publishers a revenue stream while giving AI developers scalable access to licensed material. Under PCM, publishers define usage terms and retain ownership and editorial independence, while being paid based on delivered value. The marketplace includes usage-based reporting to help publishers assess how their content is utilised. Participation is voluntary and designed to avoid complex one-to-one licensing agreements. Microsoft says the system will support organisations of all sizes, from large media houses to independent publishers.
Harder hiring
New research from LinkedIn shows that 74% of recruiters in India are struggling to find qualified candidates, despite hiring levels remaining well above pre-pandemic norms. The study attributes the challenge to a widening gap between application volume and candidate quality. More than half of recruiters (53%) say AI-generated applications have made hiring more difficult, while 47% cite shortages in critical skills. Nearly 48% report added friction in distinguishing genuine applicants from low-quality submissions.