10 articles — 🔴 5 critical , 🟡 5 important
OpenAI introduced on May 7, 2026, three new realtime voice models in the API: GPT-Realtime-2 with GPT-5-class reasoning and a 128,000-token context, GPT-Realtime-Translate that translates from 70+ input languages into 13 output languages, and GPT-Realtime-Whisper for live speech transcription.
Gemini 3.1 Flash-Lite has been generally available through the Gemini API since May 7, 2026, as a stable production endpoint. The model is optimized for speed, scale, and cost efficiency, and the preview version will be discontinued on May 25, 2026.
Google DeepMind has published the first report on the industrial impact of the AlphaEvolve agent and opened its commercial availability through Google Cloud. Klarna doubled the training speed of its transformer model, FM Logistic achieved 10.4% better routing efficiency, and Schrödinger reached a 4× speedup in molecular simulations.
AWS has released Amazon Bedrock AgentCore Payments in preview — the first managed infrastructure for autonomous AI agent payments. The service uses the open x402 HTTP protocol and stablecoins through a partnership with Coinbase and Stripe (via Privy), with budget limits and full auditability through the AgentCore console.
Researchers introduced LongSeeker — a long-horizon search agent using the Context-ReAct framework with five dynamic context management operations. The model achieves 61.5% on the BrowseComp benchmark, outperforming Tongyi DeepResearch by 18 percentage points.
A CNCF benchmark compares RAG, hybrid and pure local retrieval for AI agents fixing real Kubernetes bugs. RAG is fastest (1m16s), but the key bottleneck is not speed or cost — it is 'scope discovery', the agent's ability to recognise all affected code across multiple files.
OpenAI is expanding the Trusted Access for Cyber (TAC) program to thousands of verified defensive researchers and hundreds of teams protecting critical software infrastructure. The program introduces GPT-5.5 with reduced restrictions, and the specialized GPT-5.5-Cyber for reverse engineering and malicious software analysis.
A paper accepted at ICML 2026 introduces SQSD — a method for quantifying the contribution of individual samples to safety degradation during model fine-tuning. Researchers show that even seemingly benign fine-tuning samples cumulatively shift parameters toward 'danger-aligned' directions.
