AI news digest — April 16, 2026
10 items, each with its source.
OpenAI Announces GPT-5 General Availability and API Pricing
OpenAI has officially launched GPT-5 to all Plus users, introducing native multimodal reasoning and a 50% reduction in API costs compared to GPT-4. The model demonstrates unprecedented capabilities in long-horizon planning tasks.
openai.comEU AI Act Enforcement Begins With First Major Fines
The European Commission has issued its first set of fines under the fully enacted AI Act. Three major tech firms were penalized for failing to provide adequate transparency reports on their foundation models.
reuters.comDeepMind Unveils AlphaFold 4 With Dynamic Protein Simulation
Google DeepMind published a breakthrough paper in Nature detailing AlphaFold 4. The new model moves beyond static structures to predict full dynamic protein folding pathways in real-time.
nature.comNVIDIA Unveils Rubin Architecture GPUs for Next-Gen AI
Jensen Huang revealed NVIDIA's new Rubin GPU architecture during a special keynote. The hardware promises a 4x increase in training efficiency over the Blackwell generation and features advanced liquid cooling integration.
cnbc.comFigure AI Deploys 1,000 Humanoid Robots in BMW Plants
Figure AI announced the successful deployment of 1,000 Figure 02 humanoid robots across BMW's manufacturing facilities. This milestone marks the largest commercial deployment of autonomous humanoids in the automotive sector to date.
techcrunch.comApple Integrates On-Device LLM Into All New MacBooks
Apple has announced that its upcoming M5 MacBooks will feature a dedicated 30-billion parameter on-device language model. The system is completely localized for maximum privacy and zero-latency processing.
bloomberg.comMistral AI Raises $2 Billion Series D at $15B Valuation
French AI startup Mistral has secured a massive $2 billion funding round led by Andreessen Horowitz. The capital injection aims to expand its enterprise open-source model offerings globally and challenge proprietary competitors.
ft.comCoalition of Artists Wins Landmark Copyright Lawsuit Against AI
A federal judge ruled in favor of a coalition of digital artists in a major copyright dispute. The ruling establishes that generative AI companies must obtain explicit opt-in consent before using copyrighted works in training datasets.
theverge.comMIT Researchers Achieve Breakthrough in Q-Bit Error Correction
A new paper from MIT demonstrates a novel quantum error correction method that significantly stabilizes quantum neural networks. This breakthrough paves the way for practical quantum machine learning applications in the near future.
arxiv.orgBoston Dynamics Retires Atlas, Introduces Consumer Quadruped Robot
Boston Dynamics has officially retired its Atlas research program to focus on consumer applications. The company introduced a new, affordable quadruped robot designed specifically for household assistance and eldercare.
spectrum.ieee.orgFeed, daily deep-dive and bytes — readable offline, with push alerts for the topics you follow.