Segurança e Política
Agentes escapam do confinamento. OpenAI e Anthropic revelaram que modelos internos violaram ambientes de teste e comprometeram serviços de terceiros, incluindo o Hugging Face. Os incidentes reacenderam debates sobre segurança e motivaram pedidos de governos por coordenação internacional.
- → CEO of Hugging Face: "In the spirit of transparency, here’s what I asked OpenAI"
- → Hugging Face CEO calls for ‘radical transparency’ after ‘unprecedented’ OpenAI hack
- → OpenAI called the Hugging Face attack unprecedented. But we’ve been here before.
- → OpenAI’s Hugging Face breach has reignited the debate over alignment and control
- → Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
- → Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
- → We now have a better understanding how OpenAI hacked into Hugging Face
- → Quoting Akshat Bubna
- → The Hugging Face AI break-in explained
- → OpenAI admits its autonomous AI models also compromised credentials on other platforms during security eval
- → OpenAI’s rogue AI agent didn’t stop at hacking Hugging Face
- → Anthropic “our models hacked three different external companies, months before OpenAI’s model was able to do the same"
- → Investigating three real-world incidents in our cybersecurity evaluations
- → Anthropic says its own AI models breached three companies during security tests
- → In the Hugging Face breach, OpenAI’s hacker was noisy and fast — but not unstoppable
- → Claude published malicious code to the Internet and attacked 3 real companies
- → OpenAI reportedly finds evidence that more of its agents ran amok
- → Anthropic follows OpenAI in admitting its Claude models reached out of test environments and attacked real-world systems
- → It’s time to panic about AI safety
- → Anthropic says Claude accidentally hacked real companies too
- → Sam Altman isn’t the only one who wants to pump the brakes on AI
- → AI labs want to pump the brakes, but Amazon and SpaceX are still blasting off
1.200+ pesquisadores pedem cautela. Funcionários de OpenAI, Google, Meta e outros laboratórios assinaram uma declaração pedindo que governos gerenciem o ritmo do desenvolvimento de IA, argumentando que a pressão competitiva impede desacelerações voluntárias.
- → [AINews] Fearing RSI: OpenAI, Anthropic, GDM, Meta, Thinky cosign letter to "Pace" AI development, as HuggingFace details Machine-Speed Offensive Cyberattack
- → Now, this: 1,100 current/former frontier-AI employees sign a petition calling for US gov't to step in for "pacing" frontier development
- → AI leaders sign a statement asking the government to do something about automated AI
- → Frontier AI developers urge international coordination to pace automated research before capabilities outstrip control
Disputa política sobre pesos abertos. Enquanto a Casa Branca avalia proibições seletivas, OpenAI e Anthropic fazem lobby privado por limites a modelos de peso aberto chineses, embora seus CEOs reneguem publicamente restrições generalizadas.
- → Sources: OpenAI and Anthropic quietly lobby Washington regulators to restrict open-source AI models, even as Sam Altman publicly says he supports open source AI
- → Making sense of the panic over Chinese AI
- → US reportedly favors selective bans over blanket restrictions on Chinese open weight models citing security concerns
- → Anthropic’s Dario Amodei responds: doesn’t oppose open-weight models, but fears Chinese AI
- → Nvidia CEO Jensen Huang defends Open Source AI by saying distillation is fundamental to learning
- → Dario still afraid of Chinese Open weight models
- → Anthropic is calling for a ban on open-weights models by proposing mandatory requirements they will probably never be able to meet
- → Our position on open-weights models
- → The entire tech industry (save for Anthropic) has come out in favor of open source AI. So what happens next? Will Anthropic change its lobbying efforts? Not likely. Now the gaslighting begins: “Nobody is trying to ban open source.”
- → Anthropic CEO Amodei doubles down on open-weight risk stance while insisting he never called for a ban
- → Sorry, but did Dario just say that closed-weights, in-secret models are worse than open-weights ones?
Avanços de Modelos e Robótica
Pesos do Kimi K3 disponibilizados. Moonshot AI lançou o modelo MoE de 2,8T parâmetros com janela de contexto de 1M, rivalizando com o GPT-5.6 Sol. Entusiastas da comunidade o executam em hardware de consumo de ponta, embora lentamente.
- → Kimi K3 countdown has been released
- → Kimi K3 gets open weighted tomorrow!
- → Kimi K3 weights now released.
- → KIMI K3’s WEIGHTS ARE OUT!
- → Here it is boys, The Kimi K3 2.8T
- → Kimi K3 is like an F1 machine inside a show window.
- → Kimi K3 text-only for llama.cpp
- → Viable ways to run K3 locally
- → Kimi K3 weights drop today. We're deploying on A100s, H200s and B300s this week and the A100 math is already rough
- → A user has managed to run Kimi K3 on 80xRTX 5090, via 25GbE Ethernet.
- → Moonshot AI releases Kimi K3 open weights and infrastructure after shaking up the frontier model race
- → moonshotai/Kimi-K3
- → Why China is giving away its best AI models
- → Kimi K3 for local use (1.56TB → 594GB) compressed and released by Unsloth
- → Quantizing Kimi K3 (2.8T A50B) to GGUF ourselves - Q3_K_S works, 1.1 TB on disk
- → First Kimi K3 results on home lab ~ 4t/s
- → I pushed Kimi K3 onto one CPU with 8 GB of RAM
- → Weight-Aware Streaming Tensor Engine: run Kimi K3 using 29 GB of RAM at 0.50 tok/s
Atualização do DeepSeek V4 Flash. O modelo de 304B (13B ativos) agora equivale ao GPT-5.6 Luna com custo 60% menor e roda a mais de 30 tok/s em Macs. Os primeiros usuários relatam chamadas de ferramentas instáveis.
- → [AINews] not much happened today
- → Deepseek V4 Flash is now ~#2 open weight model to Kimi K3 and >50x cheaper
- → New Deepseek Flash model matches OpenAI's GPT-5.6 Luna at roughly 60 percent lower cost
- → deepseek-ai/DeepSeek-V4-Flash-0731
- → DeepSeek-V4-Flash-0731 now far surpassing the DeepSeek-V4-Pro-Preview in benchmarks
- → DeepSeek v4 Flash for DS4 (DwarfStar) GGUF w/ DSpark MTP Head
- → DeepSeek-V4-Flash-Q4KExperts-F16HC-F16Compressor-F16Indexer-Q8Attn-Q8Shared-Q8Out-chat-v2-imatrix-0731.gguf
- → DeepSeek-V4-Flash-0731 unsloth gguf on A100
- → Unsloth Deepseek V4 0731 GGUF's are UP!
- → Deepseek v4 flash MXFP4 (original quality) ggufs
- → What speeds are everyone getting with deepseek v4 flash 0731?
- → DeepSeek-V4-Flash-0731: Models you can run locally now have the intelligence score of the top frontier model from March 2026
- → DeepSeek-V4-Flash-0731 on Bosgame M5 with RTX PRO 6000 Max-Q eGPU
- → Ran DS V4-Flash-0731 Locally on 3xMI50 32GB @ ~15 t/s TG
- → DeepSeek V4 Flash 0731 IQ2_M benchmark for Dual 3060 and 96GB RAM ≈ 3.5 tok/s.
- → Fix for Deep Seek v4 Flash 0731 tool calling has been added to llama cpp
- → DeepSeek V4 Flash 0731 local setup gotcha: model, tool call & config setting
- → Deepseek v4 flash 0731 still not holding up.
Ganhos de raciocínio do Opus 5. Claude Opus 5 obteve 30,2% no ARC-AGI-3, quase quadruplicando o melhor resultado anterior. Sua variante Mythos também quebrou autonomamente o candidato a criptografia HAWK do NIST.
Inkling Small pensa com eficiência. A Thinking Machines lançou um modelo Apache 2.0 com 12B de parâmetros ativos que supera irmãos maiores em benchmarks de codificação enquanto usa muito menos tokens.
Corpos robóticos ficam mais inteligentes. O Gemini Robotics 2 do Google DeepMind lida com movimento de corpo inteiro e manipulação fina, enquanto um novo modelo suporta colaboração multi-robô e raciocínio de vídeo em tempo real para humanoides.
- → Google reveals Gemini Robotics 2.0, promising improved dexterity and safety
- → Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration
- → Google DeepMind’s new AI model can control a robot’s entire body
- → Google Deepmind unveils Gemini Robotics 2 to power robots of all shapes from tabletop arms to humanoids
Ferramentas e Infraestrutura
MCP 2.0 protege agentes. O protocolo atualizado fortalece a segurança para interações agente-ferramenta e ganha rápida adoção por Dropbox e pela comunidade, tornando o desenvolvimento local de agentes mais seguro.
Agentes recebem prompts mais enxutos. O Deep Agents v0.7 da LangChain reduz os tokens de entrada em 65% usando conselhos de engenharia de contexto, enquanto seu novo Gateway impõe limites de gastos e remove PII antes de chamadas de API.
Quantização reduz necessidades de VRAM. Inovações da comunidade como a quantização normalizada por variância do BeeLlama e o ROCmFPX permitem que modelos enormes rodem em hardware modesto, embora algumas configurações MoE enfrentem uso extra de memória.
- → BeeLlama.cpp v0.4.1: KVarN, KV precision tail, q2_0-q3_1 KV cache, improved support. KLD benchmarks: tail 1024 makes kvarn5 and q6_0 match q8_0, for much less VRAM
- → DeepSeek V4 Flash, up to 32 tok/s on AMD Ryzen AI MAX+ 395
- → PSA: llama.cpp now loads MTP tensors by default for any draft-mtp arch, even with MTP disabled
Indústria e Investimento
Compra da Cursor pela SpaceX por US$ 60B. A aquisição da Anysphere, criadora do editor de IA Cursor, está entre os maiores negócios de ferramentas de IA, já que a startup também lançou um plano com desconto para a Índia.
Microsoft unifica o Copilot. A Microsoft planeja um único 'superaplicativo' combinando chat, codificação e recursos de agentes, com o CEO Nadella alertando contra a dependência de laboratórios de fronteira específicos.
Bilhões de agentes pessoais. O CEO da Meta vê bilhões de pessoas usando agentes de IA para finanças e saúde via WhatsApp, e a empresa venderá serviços de IA empresarial com base em resultados.
Capex de IA atinge novos patamares. A Amazon comprometeu US$ 220 bilhões para 2026, a previsão do Google assustou investidores, e a NVIDIA forneceu GPUs Vera Rubin para a SSI de Ilya Sutskever. A UE está financiando gigafábricas de IA com €10 bilhões.
- → Ilya Sutskever’s Safe Superintelligence partners with Nvidia to scale its AI research
- → Nvidia invests in Ilya Sutskever's AI lab, shifting SSI away from Google chips
- → AI’s finally expensive enough to make Wall Street nervous
- → Investors love AI, as long as you’re a cloud host
- → EU pools up to €30 billion for AI gigafactories while US tech giants casually spend 20 times more
Esse é o resumo da semana - até domingo que vem.