All news with #mistral ai tag

🎙️ AWS added Voxtral-Mini-4B-Realtime-2602 to Amazon SageMaker JumpStart, a multilingual, low-latency speech-transcription model from Mistral AI. The model offers natively streaming architecture for real-time transcription across 13 languages and configurable delay/accuracy trade-offs. Customers can deploy it via the SageMaker Studio Models section or the SageMaker Python SDK for rapid integration into speech applications.

🧭 AWS now offers Ministral-3-14B-Instruct-2512 in Amazon SageMaker JumpStart, adding a compact multimodal foundation model optimized for edge deployment. The 14B-parameter model supports image analysis, agentic workflows with native function calling and JSON output, and multilingual understanding across dozens of languages. Customers can deploy the model from SageMaker Studio or via the SageMaker Python SDK with a few clicks to build AI assistants, agentic systems, and vision-enabled applications on AWS.

🔒 Mistral AI says the TeamPCP group is offering nearly 450 repositories allegedly stolen from the company’s codebase, demanding a $25,000 buy‑it‑now price and threatening to leak the files within a week if unsold. The hackers claim about 5 gigabytes of internal source code used for training, fine‑tuning, benchmarking, model delivery, and inference was exfiltrated after a compromise tied to the Mini Shai-Hulud supply‑chain attack and tampered TanStack packages. Mistral confirmed some SDK packages were contaminated briefly but says forensic analysis found no compromise of core repositories, hosted services, or managed user data.

🛡️ The TeamPCP group compromised 170 npm and PyPI packages on May 11, rapidly spreading malicious code across ecosystems including the @tanstack router and Mistral AI SDKs. Attackers abused GitHub Actions' pull_request_target trigger to harvest OIDC tokens and inject the Mini Shai-Hulud malware, which steals credentials and carries a destructive dead-man’s switch. Security vendors detected the compromise quickly; affected users should check lockfiles, pin known-good versions, and rotate exposed credentials.

⚠ TeamPCP's "Mini Shai-Hulud" campaign has trojanized npm and PyPI packages from maintainers including TanStack, Mistral AI, OpenSearch, UiPath, and Guardrails AI, deploying an obfuscated credential stealer that targets cloud services, crypto wallets, AI tools, messaging apps and CI systems. The malware exfiltrates data via a Session Protocol domain (filev2.getsession[.]org), a typosquat domain and GitHub API dead-drops, and persists through IDE hooks in Claude Code and VS Code. Attackers abused GitHub Actions OIDC permissions and produced malicious packages with valid SLSA attestations; TanStack's cluster was assigned CVE-2026-45321 (CVSS 9.6).

📄 Mistral Document AI 2512 in Microsoft Foundry combines high-end OCR (mistral-ocr-2512) with contextual extraction (mistral-small-2506) to convert scans, photos and digital documents into structured JSON and markup while preserving layout, tables and handwritten notes. It emphasizes enterprise-grade accuracy, multilingual coverage and private/secure inference. Paired with the ARGUS accelerator, organizations can deploy end-to-end pipelines quickly and switch OCR providers at runtime.

🚀 Amazon Bedrock now offers Mistral AI Devstral 2 123B, an open-weight 123B-parameter LLM optimized for agentic software engineering workflows. The model focuses on code generation, automation, and reliable multi-step reasoning, supporting long-context comprehension for multi-turn coding tasks. Bedrock exposes Devstral 2 via a single, fully managed API so customers do not need to provision infrastructure or host models. It is intended for production coding assistants, automated code review, and complex software development agents and is available in select AWS Regions.

🚀 Amazon Web Services announced that Amazon Bedrock now supports the latest open-weight models in Asia Pacific (Sydney) through the bedrock-mantle endpoint. The update brings models from providers including DeepSeek, Google, MiniMax, Mistral, Moonshot AI, Nvidia, and OpenAI, expanding local model choice. Powered by Project Mantle, bedrock-mantle delivers a distributed, serverless inference engine with advanced quality-of-service controls, automated capacity management and unified pools. It also offers out-of-the-box OpenAI API compatibility to simplify integration for developers.

🔍 Threat actors have been probing misconfigured proxy servers to access paid large language model (LLM) endpoints, generating over 80,000 sessions since late December, according to GreyNoise. Attackers used low-noise queries to fingerprint models without triggering alerts and targeted vendors such as OpenAI, Anthropic, Google, Meta, Mistral and others. While GreyNoise reports no observed exploitation or data theft, the scale of enumeration indicates reconnaissance with possible malicious intent. Recommended mitigations include restricting Ollama model pulls to trusted registries, applying egress filtering, blocking known OAST callback domains at DNS, rate-limiting suspicious ASNs, and monitoring JA4 fingerprints.

🚀 Microsoft has added Mistral Large 3 to Foundry on Azure, offering a high-capability, Apache 2.0–licensed open-weight model optimized for production workloads. The model focuses on reliable instruction following, extended-context comprehension, strong multimodal reasoning, and reduced hallucination for enterprise scenarios. Foundry packages unified governance, observability, and agent-ready tooling, and allows weight export for hybrid or on-prem deployment.

🚀 Amazon Bedrock now offers Mistral Large 3 and the Ministral 3 family alongside additional Mistral AI checkpoints, giving customers early access to open-weight multimodal models. Mistral Large 3 employs a granular Mixture-of-Experts architecture with 41B active and 675B total parameters and supports a 256K context window for long-form comprehension and agentic workflows. The Ministral 3 series (14B, 8B, 3B) plus Voxtral and Magistral small models let developers choose scales optimized for production assistants, RAG systems, single-GPU edge deployment, or low-resource environments.

🔎 Microsoft researchers disclosed a new side-channel attack called Whisper Leak that can infer the topic of encrypted conversations with language models by observing network metadata such as packet sizes and timings. The technique exploits streaming LLM responses that emit tokens incrementally, leaking size and timing patterns even under TLS. Vendors including OpenAI, Microsoft Azure, and Mistral implemented mitigations such as random-length padding and obfuscation parameters to reduce the effectiveness of the attack.

🔒 Microsoft has disclosed a novel side-channel called Whisper Leak that can let a passive observer infer the topic of conversations with streaming language models by analyzing encrypted packet sizes and timings. Researchers at Microsoft (Bar Or, McDonald and the Defender team) demonstrate classifiers that distinguish targeted topics from background traffic with high accuracy across vendors including OpenAI, Mistral and xAI. Providers have deployed mitigations such as random-length response padding; Microsoft recommends avoiding sensitive topics on untrusted networks, using VPNs, or preferring non-streaming models and providers that implemented fixes.

🔍 Microsoft researchers disclosed "Whisper Leak", a new side-channel that can infer conversation topics from encrypted, streamed language model responses by analyzing packet sizes and timings. The study demonstrates high classifier accuracy on a proof-of-concept sensitive topic and shows risk increases with more training data or repeated interactions. Industry partners including OpenAI, Mistral, Microsoft Azure, and xAI implemented streaming obfuscation mitigations that Microsoft validated as substantially reducing practical risk.

🔐 Google Cloud’s Vertex AI now supports secure self-deployment of proprietary third-party models directly into customer VPCs via the Model Garden. Customers can discover, license, and deploy closed-source and restricted-license models from partners such as AI21 Labs, Mistral AI, Qodo and others, with one-click provisioning and managed inference. Deployments adhere to VPC-SC controls, selectable regions, autoscaling, and pay-as-you-go billing. This central catalog brings Google, open, and partner models together for enterprise-grade control and compliance.