Perplexity Secures 650 Million Series D at 9 Billion Valuation
Perplexity closed a 650 million dollar Series D round on July 22 2026 led by a16z at a 9 billion post-money valuation. The AI search startup reported 180 million dollars in annualized revenue and 42 million monthly active users. Funds will expand enterprise search products and data center capacity in Europe.
Mistral AI Unveils Large 3 Model with 4M Token Context
Mistral AI released its Large 3 model on July 21 2026 featuring a 4 million token context window and improved multilingual performance. The model scores 92.4 on MMLU and supports native tool calling for agent workflows. This positions Mistral as a stronger competitor to US frontier labs in enterprise deployments.
Google DeepMind and Novartis Extend AI Drug Discovery Partnership
Google DeepMind and Novartis extended their AI drug discovery collaboration on July 17 2026 for five additional years. The new agreement includes joint development of a 1 billion parameter protein design model. Initial candidates from the partnership enter Phase 1 trials in 2027.
Hugging Face Releases Transformers 5.0 with Native Agent Support
Hugging Face launched Transformers 5.0 on July 19 2026 adding native support for building and deploying AI agents. The update includes a new Agent class and integration with 12 major model providers. Over 1.2 million developers have already downloaded the beta.
EU Commission Proposes Mandatory AI Model Audits Starting 2027
The European Commission published draft rules on July 18 2026 requiring independent audits of foundation models above 10^25 FLOPs. Audits must assess bias, safety, and copyright compliance. The proposal faces parliamentary vote in October.
Stanford Team Achieves 40 Percent Improvement in LLM Reasoning Efficiency
Stanford researchers published a new training method on July 20 2026 that improves LLM reasoning accuracy by 40 percent while using 35 percent less compute. The technique called Recursive Preference Optimization was tested on 7B to 70B models. It will be open sourced next week.
Apple Integrates On-Device AI Agents into iOS 20 Beta
Apple seeded iOS 20 beta 3 on July 24 2026 with on-device AI agents capable of completing multi-step tasks across apps. The agents run entirely on A19 silicon with no cloud calls for core functions. Early developer feedback highlights improved Siri reliability.
LockBit Affiliates Breach 1.9 Million Records from European Hospital Network
LockBit affiliates claimed responsibility for a July 19 2026 breach of a 12-hospital network in Germany exposing 1.9 million patient records. The group demanded 12 million euros in ransom. German authorities confirmed the incident on July 21.
Cohere Raises 800 Million Series E at 8.5 Billion Valuation
Cohere closed an 800 million dollar Series E round on July 23 2026 at an 8.5 billion dollar valuation. Inovia Capital and Tiger Global led the round with participation from existing investors. The funding will expand enterprise deployment of Command R+ across Europe and Asia.
Meta Unveils Llama 4 with 10M Token Context and Native Video
Meta released Llama 4 on July 22 2026 featuring a 10 million token context window and native video understanding. The model scores 92.4 percent on MMLU and 89.1 percent on VideoMME benchmarks. It positions Meta as a direct competitor to closed frontier labs in multimodal reasoning.
Microsoft Integrates Phi-4 Model into Windows 12 Copilot+ PCs
Microsoft began rolling out its Phi-4 3.8 billion parameter model to Windows 12 Copilot+ devices on July 19. The on-device model delivers 42 tokens per second on Snapdragon X Elite chips. It supports offline code completion and image editing without cloud calls.
Ransomware Group Qilin Exfiltrates 2.8 Million Records from UnitedHealth
Qilin ransomware operators claimed theft of 2.8 million patient records from UnitedHealth Group on July 20. The group demands 50 million dollars in Bitcoin. UnitedHealth confirmed the intrusion began on July 12 via a compromised VPN credential.