All sections
News
451 stories · page 20 of 51
SecurityThree holes in the Linux kernel went onto the exploited list, and their fixes are months oldCISA added three kernel vulnerabilities already used in attacks on 18 September, with a deadline of 21 September. Their NVD entries date from September and October 2025 and June 2026. The problem is the machines that never got the fix.AudioGrok Voice Transcribe 2.0 transcribes twice as accurately as the first version in SpaceXAI tests, at the same priceThe new speech-to-text model is built on the audio model behind Grok Voice, detects the language itself and follows a switch mid-recording. The price stays at $0.10 per hour of audio for batch and $0.20 for streaming.BuildersGoogle closed Gemini 2.5 to new usersIn a note dated 18 September Google limited access to the Gemini 2.5 models in the API to users who have actively used them. They are not deprecated, but for a new app Google points to 3.5 Flash-Lite or 3.8 Flash, citing capacity.ControlBrussels proposed: no social media under 13, your own account from 15, AI companions off by defaultThe Commission adopted the draft EU KIDS Act. Platforms will have to prove for themselves that they are safe for children. It is a proposal and now goes to Parliament and the Council.ScienceClaude sped up more than 30 biology models by about 4 times on average in under four weeks and opened the codeAn internal Anthropic research model, supervised by two people with no background in optimisation, optimised the open models for protein structure and design. Plus a new mode that predicts systems over 10,000 tokens on a single server.ControlAnthropic opened Claude to verified biologists with fewer safeguards, and for risky projects without the biology onesThe life sciences verification programme gives approved teams more permissive classifiers. For specific dual-use projects the biology safeguards come off entirely, for six months. The price is 30 days of data retention for monitoring.RoboticsFigure's humanoid worked in 30 unfamiliar homes without a single day of training there and succeeded in 56 per cent of attemptsHelix 2.5 is pretrained on a vast dataset of human behaviour, and from one model it yields three skills: tidying a living room, folding towels and making a bed. Per Figure, pretraining raises success in an unfamiliar home from 9 to 56 per cent.FrontierOpenAI turned GPT-6 Astra into a lawyer: 54 instead of 38.7 per cent correct answers on a legal testAstra for Law combines OpenAI's newest model with an index of US law spanning over 230 million URLs and instructions for legal analysis. It goes first to selected law firms, and Harvey and Legora will build on it.
451 stories