Signal Over Noise.

Telecommunications giant AT&T is integrating OpenAI's generative AI to support its in-house legal operations, a significant move for enterprise AI adoption in specialized professional fields.

Intermittent connectivity issues with OpenAI's services highlight growing infrastructure challenges and the critical need for more robust, standardized access protocols as AI platforms scale to meet global demand.

IBM Research has released Granite Timeseries Ensemble, an open framework that combines four pre-trained models to simplify and improve the robustness of time-series forecasting for enterprise applications.

NVIDIA introduces Kumo Relational, a pre-trained foundation model that outperforms supervised ML on relational databases for tasks like churn and fraud prediction without requiring any model training.

Chinese AI lab DeepSeek reportedly secures a massive $15 billion funding round from investors including CATL and Tencent, just days after facing US government restrictions on its technology.

A Blackstone executive addresses growing market concerns about the massive debt financing the AI infrastructure buildout, arguing the risks are contained by high-quality customers.

Anthropic signals its public market ambitions with a confidential S-1 filing, a $65B funding round, and a rapid succession of new Claude model releases aimed at enterprise adoption.

OpenAI and Google Cloud announce a new partnership to offer OpenAI's generative AI models, including GPT-4, directly on the Google Cloud Platform via Vertex AI.

A franchisee reports a 94% reduction in cash reconciliation time using a 'ChatGPT Work' solution, signaling a shift for OpenAI's technology toward specialized back-office financial automation.

NVIDIA has released Kumo Forecast, an open-source time-series forecasting model that combines a pretrained Transformer backbone with a novel historical pattern retrieval system and a built-in interpretability framework.

The Allen Institute for AI's new FairStress benchmark reveals that LLMs exhibit hidden biases by changing correct, evidence-based decisions more readily when pressured to favor minority-coded groups.

NVIDIA and the University of Washington release RFD4-Proteina, an open model that accelerates protein design with a novel reward-guided inference-time search, available as an open model and a NIM microservice.

NVIDIA announces Sentry, a new security framework running on BlueField-4 DPUs that provides hardware-isolated runtime protection and policy enforcement for AI agents.

AI safety lab Anthropic has drafted a legislative proposal for Congress that would block the release of an AI model if it poses a 'catastrophic risk,' defined as the potential for over 50 deaths or $1 billion in damages.

OpenAI's initiative to advance coding in low-resource languages is being partially obscured by its own website's stringent security measures, which block automated access and raise questions about information transparency in the AI sector.

AnyLanguageModel 1.0 has been released, offering Apple developers a unified, drop-in API to integrate a wide range of local and hosted LLMs into their apps with consistent support for features like tool calling and guided generation.