Latest AI News Today: 12 November 2026

ElevenLabs has released Scribe v3 and Scribe v3 Live, speech to text models for messy recorded and real time audio. The company says Scribe v3 cuts word errors by about 25 percent compared with its predecessor across 102 languages, and Scribe v3 Live uses conversational context to tell when a speaker has finished. Independent benchmarks rank Scribe v3 among the most accurate transcription models, though it is mid pack on speed and price.

Zimbabwean author Simbarashe Kundizeza was cleared of using AI to write The Blue Fighter, which won the £10,000 Caine prize for African writers. The prize said a month-long investigation found the story was human-written, after an Atlantic article said AI-detection software Pangram had deemed it almost completely artificially generated. The review included interviews and metadata analysis of drafts, and Kundizeza said he was relieved.

NHS England deleted a claim that 137 hospital trusts were benefiting from Palantir's AI health data system after 29 said they were not using all or most of it. The deletion follows a list describing trusts as live with the federated data platform, which is meant to improve patient outcomes and speed up bed discharges. It comes as ministers weigh extending Palantir's £330m contract to 2029 and the statistics watchdog examines the claims.

Australia's government announced a law that would impose mandatory safety rules on frontier AI companies for the first time. The rules target the most advanced, high-risk models and would require independent audits, ongoing safety checks, published safety frameworks and immediate reporting of security incidents. A new oversight body would work with the Australian AI Safety Institute to enforce compliance and impose steep penalties.

KPMG's global head of consulting said AI is making the advice that underpinned consulting fees more accessible, raising client expectations and reducing the value of traditional billable hours. Firms are struggling to price work as clients want outcomes while procurement still compares staff numbers and hourly rates, and AI adds computing and model usage costs. AI is also challenging the graduate pipeline that has powered consulting firms.

Jensen Huang's alleged affair with a Chinese intelligence recruit is strengthening U.S. lawmakers' push to add chip export controls to this year's National Defense Authorization Act. The bills aim to limit advanced chip flows to China ahead of President Donald Trump's planned meeting with Xi Jinping, and the scandal has weakened Nvidia's lobbying power.

Australia's workplace relations minister Amanda Rishworth launched a best practice guide urging employers to give workers a clear plan as AI changes jobs. It says entry-level roles will be automated first and that AI could bring higher wages and shorter hours, but also risks intrusive surveillance, poor rostering and discrimination. The guide follows Qantas and Canva job cuts in AI-related restructures.

Google donated OpenRL, a reinforcement learning project from its GKE Labs team, to the Cloud Native Computing Foundation. The project, which debuted as a research preview in June 2026, offers a self-hosted training API for a company's own Kubernetes cluster. It separates the reinforcement learning loop from compute plumbing, allowing infrastructure engineers to handle scaling while researchers keep control of experiments.

A police watchdog report found that retrospective facial recognition is beset by errors and bias, and some officers acted on potential matches without enough checks. Inspectors said results are intelligence rather than evidence, and poor-quality probe images reduced reliability. The report also found the algorithm is more likely to wrongly flag ethnic minorities, raising the risk of wrongful arrests and undermining public confidence.