VentureBeat
United States · English · Digital
A technology news website covering developments in artificial intelligence, enterprise software, and tech industry trends.
- 3/day · 7-day avg
- Top beat: Technology
- Tracked since 2026
- 25 United States peers
Latest stories
- Researchers trained an open source AI search agent, Harness-1, that outperforms GPT-5.4 on recalling relevant informationResearchers developed an open source AI search agent called Harness-1 that outperforms GPT-5.4 in recalling relevant information. The project involved collaboration between the University of Illinois at Urbana-Champaign, UC Berkeley, and Chroma.
- The Agentic Reckoning: Enterprise AI organizations have a runtime problem, not a model problem — and most are building the wrong solutionEnterprise AI organizations face a runtime problem rather than a model problem, with most building the wrong solution, according to VentureBeat's Pulse Research in Q1 2026.
- When Claude changed, everything changed: Managing AI blast radius in productionThe article discusses how a change in Claude's system affected its functionality, requiring users to manually retrieve data from multiple sources, highlighting challenges in managing AI blast radius in production.
- Agentic AI solved coding — and exposed every other problem in software engineeringAgentic AI has significantly improved coding efficiency but raised concerns about other unresolved issues in software engineering.
- When Claude changed, everything changed: Managing AI blast radius in productionThe article discusses how a system that converted natural-language questions into API calls was used by analysts, account managers, and operations leads to streamline data retrieval from multiple dashboards and BI tools.
- Microsoft AI chief says company was “set free” from OpenAI to pursue superintelligenceMicrosoft's AI chief stated the company was 'set free' from OpenAI to focus on superintelligence, ending a three-year partnership that involved over $13 billion in investments and provided access to advanced AI models.
- Microsoft's AI Futurist explains how he uses Copilot — and the real-world problems enterprises are solving with agentsMicrosoft's AI Futurist discussed the use of Copilot and how enterprises are implementing agents to solve real-world problems, emphasizing the importance of context, governance, and secure access.
- AI agents are learning on the job — just not for your whole teamAI agents improve when corrected by individuals, but these improvements are not shared among team members, requiring each user to start from scratch.
- Meta's AI support agent bound recovery emails for anyone who asked. Your SOC never saw an alert.Meta's AI support agent sent recovery emails to accounts upon request without triggering security alerts, as authorized agents logged the transactions without detection.
- Anthropic says 80% of its new production code is now authored by Claude — how your enterprise can keep upAnthropic reports that over 80% of its new production code in May was written by its AI model Claude, highlighting the increasing role of AI in software development.
- Google's new open source Gemma 4 12B analyzes audio, video — and runs entirely locally on a typical 16GB enterprise laptopGoogle released Gemma 4 12B, an open-source AI model that analyzes audio and video and runs locally on a 16GB enterprise laptop.
- Enterprise AI agents keep creating data silos. Microsoft's Build answer is Microsoft IQ and Rayfin.Enterprise AI agents create data silos, and Microsoft addresses this with Microsoft IQ and Rayfin to improve data integration and governance.
- Alibaba's Qwen3.7-Plus supports text, video and imagery inputs at low cost of $0.4/$1.6 per 1M token — but it's proprietaryAlibaba released Qwen3.7-Plus, an AI model with multimodal support for text, video, and imagery at lower costs than its predecessor, but it remains proprietary.
- Perplexity AI unveils hybrid local-cloud inference system at Computex 2026Perplexity AI introduced a hybrid local-cloud inference system at Computex 2026, which autonomously decides where to process AI tasks in real time.
- The Agentic Reckoning: Enterprise AI organizations have a runtime problem, not a model problem — and most are building the wrong solutionEnterprise AI organizations face a runtime problem rather than a model problem, with most building the wrong solutions, according to VentureBeat's Pulse Research in Q1 2026.