ElevenLabs · Original source
ElevenLabs has described its work with Dutch telecoms provider KPN on Mobiel AI, a voice AI add-on for KPN's small-business mobile subscriptions. The service offers voicemail transcription, call summaries with follow-up actions, and a call menu that routes callers to colleagues. KPN chose ElevenLabs for Dutch speech quality, and the service runs on KPN's network and data centres so voice data stays in the Netherlands.
Google · Original source
Google is investing more than $6 million in new research grants and capacity-building initiatives at the National University of Singapore and Nanyang Technological University. Grants include $1.5 million for Alzheimer's research at NUS, funding for computing efficiency and regional languages, AI safety studies at NTU and a $750,000 AI for Science capacity fund at NTU. Google.org and Google DeepMind are also providing $3 million to INSEAD's Thrive Lab in Singapore.
Microsoft · Original source
Lumen Technologies and Microsoft have expanded their collaboration to jointly sell cloud networking to enterprises adopting AI on Azure. The companies plan joint account planning where network modernisation, Azure migration and AI projects overlap, and Lumen intends to bring its cloud networking and application connectivity capabilities to Azure customers. Alkira, a Lumen Connect solution, is available through Microsoft Marketplace. Further details will follow as go-to-market plans progress.
Salesforce · Original source
Salesforce has signed a definitive agreement to acquire Cresta, which makes voice AI agents that handle complex customer calls and tools that guide human service representatives in real time. Cresta's customers include Marriott, Comcast and Alaska Airlines. Salesforce plans to connect Cresta's voice agents and conversation insights with its customer data and Agentforce platform. The deal is expected to close in the first quarter of Salesforce's fiscal year 2028, subject to regulatory approvals.
OpenAI · Original source
OpenAI is planning its next internal security sprint, partnering with Cathedral, whose researchers will use OpenAI's models alongside its teams to test its defenses. OpenAI says its last sprint dedicated 25% of its product engineers to finding and fixing weaknesses using Codex. The company also cites $1 billion in subsidized access over six months for frontline defenders through Daybreak, and bigger bug bounty rewards.
OpenAI · Original source
OpenAI is raising rewards in its Security Bug Bounty program, more than quadrupling the midpoint of reward ranges for medium, high and critical findings on average. A critical vulnerability in ChatGPT, Codex or the API Platform can now earn up to $30,000, and exceptional findings up to $172,000 with bonuses. New incentives include quality multipliers, a five-tier loyalty program and priority campaign bonuses, with retroactive payments for recent awards.
OpenAI · Original source
OpenAI has described Pando, a Git-compatible backend it built for its monorepo to handle the volume of concurrent work enabled by Codex. Pando coordinates updates to each branch independently and reuses stored packfiles across fetches, served from regional caches. OpenAI says it built the Rust-based system with Codex in 10 weeks, that it cut median push latency by nearly 80%, and that it delivers about a petabyte of Git data per day. OpenAI is contributing related improvements back to upstream Git.
OpenAI · Original source
OpenAI has published a customer story on Qodo, which builds AI code review and software quality tools and offers a plugin in the Codex marketplace. The plugin gives Codex codebase knowledge, organizational rules and independent review, aiming to catch P0 bugs across customer environments of 10,000 or more repositories before a pull request is opened. Qodo uses OpenAI models including GPT-6 Astra and sees an opportunity in OpenAI's Decisions API to select which rules apply.
OpenAI · Original source
OpenAI has introduced LifeVal, a benchmark measuring how well AI models help people complete everyday tasks such as finding childcare, managing finances and handling government paperwork. Its tasks draw on the American Time Use Survey, the Multinational Time Use Study and a new demographically representative survey of 3,995 respondents. OpenAI says its latest models score very highly on many LifeVal tasks, with published examples scoring between 90.4% and 100%.
xAI · Original source
xAI has described Data Bot, a Shared Bot in Grok Bot that the SpaceXSI data science team built to answer colleagues' data questions. It queries the company's data warehouse, runs analyses and experiment readouts, builds charts and Databricks or Hex dashboards, and can join Slack channels to post scheduled reports. The bot draws on skills and a reviewed shared memory maintained by the data team, and handled an estimated 70 hours of work last week.
Sakana AI · Original source
Sakana AI has published a report on its 26 October symposium in Tokyo, where it announced it is moving into pre-training its own models. Chief Scientist Takuya Akiba introduced Sakana Sawara (provisional name), an agent-native language model series built from scratch, with pre-training of Sawara Small due to start in December and weights planned for release next spring. Chief Scientific Advisor Jürgen Schmidhuber gave the keynote on world models and Physical AI.
OpenAI · Original source
OpenAI has launched the Grid Protectors Program, under which its future compute campuses will cut electricity demand when local grids are under strain. Its planned Effingham County campus in Georgia is the pilot, with an agreement allowing up to 1,000 megawatts of flexible demand response with Georgia Power. OpenAI says it will pay the full infrastructure and electric-service costs of the project, and Georgia Power projects about $950 million in annual customer savings from 2029 from its large-load portfolio.