SpaceXAI Releases Grok Voice Transcribe 2.0, Claiming 2x Accuracy at Same Price
SpaceXAI launched Grok Voice Transcribe 2.0, claiming twice the accuracy of version 1.0 at the same price.
Latest news, developments and reporting on Large Language Models.
SpaceXAI launched Grok Voice Transcribe 2.0, claiming twice the accuracy of version 1.0 at the same price.
Apple's iPhone 18 Pro series went on sale in China on Sept. 18, with Pro Max models drawing resale premiums and Pro models breaking issue price. OpenAI also reopened its $200 ChatGPT Pro tier to eligible former subscribers.
Google confirmed Gemini accessed three real companies during a May cybersecurity test after a bug exposed the model to the internet, the first such disclosure by the company.
Jacob Coxon resigned from Anthropic and warned that OpenAI and Anthropic are not acting responsibly, as other AI insiders raise concerns about the pace of development and safety.
Decision Tree Research CEO Greg Allen says AI has already produced serious incidents and warns model releases are outrunning safety technologies, according to Bloomberg.
Ilya Sutskever says AI will keep improving and eventually do everything humans can do, in a TechRadar quote of the day.
Three Hacktron AI researchers used Anthropic's Claude to breach OpenAI's GitHub repository after exploiting flaws in OpenAI's forum and SSO system, according to a report. OpenAI patched the issue within 14 hours, and the HEIF Heist bug series also affected Slack, Meta and other firms.
Anthropic said Friday it picked Accenture as its first embedded evaluator, with Accenture's Faculty staff set to red-team models and test safeguards. The companies plan to invest at least $1 billion over five years as Anthropic advances CEO Dario Amodei's proposal for slower AI development.
Android Auto users say Gemini no longer controls smart home devices, after similar Wear OS issues.
A U.S. armed operation against a Chinese vessel was aborted at the last minute this spring after officials found the intelligence behind it had been hallucinated by an AI chatbot, CNN reported.
Sony Music Entertainment and Universal Music Group have sued AI music generator Suno again, alleging its v6 models were trained partly on unlicensed music and could infringe more than 60,000 recordings.
Anthropic opens a Bay Area wet lab using Claude robots for biology research; it says the site is not for drug discovery.
Jina AI has released jina-ocr-v1, a 3.4B-parameter MoE visual document parser with built-in speculative decoding, open weights for research and non-commercial use. It scores 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, and parses 2.57 pages per second on one A100 40 GB.
TypeSafe AI released Jev, a transformer model that outputs calibrated probabilities instead of text, drawing developer interest for cheaper, faster automation.
Apple's iOS 27 arrives with more than 20 new Apple Intelligence features, led by a Siri rebuilt for the large language model era, plus AI additions across Photos, Messages, Mail, Safari and systemwide tools, according to 9to5Mac.
Newsom orders California to explore a verified AI kill switch and audits as Congress stalls on oversight.
PrismML launched Bonsai 2 27B, a 5.9 GB ternary multimodal model based on Qwen3.8 27B that runs on consumer hardware and retains 98.2% of its average benchmark performance, according to the company.
Nvidia’s Nader Khalil and Sydney Sykes will lead a Builders Stage session at TechCrunch Disrupt 2026 in San Francisco, examining how startups should choose between open and proprietary AI models.
Hacktron used Claude to breach OpenAI employee accounts and its GitHub Monorepo in under 72 hours, The Wall Street Journal reports.
AI agents dominated the week as new products arrived alongside slowdown calls and fresh security controls.
More than 100 AI experts and evaluators urged frontier AI companies to give independent safety evaluators protected, transparent access to models and development practices.
China’s Zhunzhun Intelligent released LimiX-2, a 400M-parameter structured-data foundation model that ranks first on TabArena, TALENT and BCCO, surpassing models from Google, Amazon, SAP and others.
Security researchers at Hacktron AI used Anthropic’s Claude to breach OpenAI employee accounts via Discourse flaws, then reported it under OpenAI’s bug-bounty program. OpenAI fixed the issues and paid $6,500.
TechRadar reports that white hat hackers breached OpenAI using Anthropic's Claude in under 72 hours, describing the episode as an illustration of how fast AI capabilities are moving.