Here are today's news for 13th of October about Meta, Anthropic, Google and OpenAI. Let's dive into the details. On October 12, Meta introduced Muse Wonder and Muse Wonder Full, its strongest models yet for personal and professional AI agents. Meta says Active Memory retrieves useful details across conversations, while Active Focus lets models work through obstacles, change approach and seek clarification or permission before consequential actions. Wonder is available through Meta’s Model API and Muse Code; Wonder Full is limited to Muse Code subscribers. Wonder’s API price is $4 per million input tokens and $20 per million output tokens; Muse subscriptions cost $20 or $100 monthly, while Muse Code plans cost $50 or $200. Independent Artificial Analysis scores were 52 for Wonder and 56 for Wonder Full, versus 48 for Muse Spark 1.3; Wonder’s cost per benchmark task was $4.52. The launch puts Meta into closer competition with Anthropic and OpenAI on coding and agent work. Both models were trained at Meta’s gigawatt, scale Prometheus data center. Anthropic released Claude Haiku 5.5 on October 7, completing its 5.5 model lineup after launching two other family models in the preceding month. The company says the new Haiku is built for high, volume, latency, sensitive work, including classification, extraction and routing, and costs about 75% less to run than Haiku 4.5 on average. For prompts up to 100,000 tokens, API pricing is $0.10 per million input tokens and $0.50 per million output tokens; longer prompts cost $0.50 and $2.50, respectively. The model supports a one, million, token context window, up to 128,000 output tokens and adjustable effort, a first for Haiku. Anthropic reports 72.4% on OSWorld 2.1 computer, use testing, versus 15.7% for Haiku 4.5, and 39.2% on Terminal, Bench 4.0, versus 0%. It is available on Claude and major cloud platforms. The low price targets developers running many routine AI calls, intensifying competition for workloads where cost and speed matter more than top, tier reasoning. Google DeepMind released EmbeddingGemma 2 on October 6, 2026, an open, weight multimodal embedding model built on the Gemma 4 architecture. Its 740 million parameters map text, code, images, video and audio into one 768, dimensional vector space, enabling cross, media search on local devices. Developers can load a 270, million, parameter text, and, code configuration or add modular vision and audio encoders; Google says the text, only model can run with about 191 MB of active RAM. The Apache 2.0 license permits commercial use, with weights available through Google and Hugging Face. The model’s 8,192, token context window is four times EmbeddingGemma’s, handling up to 5.5 minutes of audio, 29 images or 58 video frames. In Google’s reported benchmarks, it scored 78.68 on MTEB Code, up from 68.76 for its predecessor, while multilingual MTEB reached 61.36 versus 61.15. On Oct. 13, OpenAI, Microsoft, Anthropic, Google and more than 100 technology, cybersecurity, finance and infrastructure organizations urged governments and industry to mount a coordinated defense against AI, enabled cyberattacks. The open letter, signed by 116 entities, warns attacks will become “far more widespread and sophisticated” in the coming months as models improve. It calls for stronger security standards, rapid repair and validation of dangerous vulnerabilities, shared threat intelligence and tested response playbooks. Signatories also want governments to fund defenses for under, resourced essential services and make defensive AI, training and hands, on support available to operators of hospitals, water systems and internet infrastructure. The appeal reflects a competitive shift: AI firms are racing to deploy capable models, while security teams fear attackers may adopt them faster than defenders. On October 8, 2026, the UK Information Commissioner’s Office said 10 AI developers, Amazon, Anthropic, Apple, Cohere, DeepSeek, Google, Meta, Microsoft, OpenAI and Stability AI, had made or committed to make changes after two years of regulatory scrutiny. The steps include clearer explanations of how personal data is used, better routes for people to exercise data rights, and stronger assessments of privacy safeguards; the ICO is monitoring delivery, not announcing fines. The changes matter to anyone whose information may be used in training or appear in model outputs, and raise the compliance bar across competing foundation, model providers. The regulator says difficult questions remain, including how sensitive data is handled and how people can seek removal after training. Alongside its findings, the ICO opened a six, week call for evidence on agentic AI, due to close November 20, and said it had made enquiries with OpenAI, Anthropic, Meta and the UK AI Security Institute after reports of agents bypassing safeguards. That's all for today's AI news briefing. Have a great day!