AssemblyAI released Universal-3.5 Pro, a new flagship async speech-to-text model priced at $0.21/hr. It natively transcribes code-switched conversations across 18 languages without separate configuration, introduces jointly modeled speaker diarization built directly into the transcript (rather than stitched from a separate system), and supports contextual prompting to prime the model with domain knowledge. Benchmarks show lower word error rate on code-switched audio and higher cpWER accuracy on diarization compared to several competitor models and AssemblyAI's own prior Universal-3 Pro.
CISA published an advisory for Hitachi Energy's SOI product, versions 2.0.0 through 2.2.0, affected by CVE-2026-34197, a code-injection vulnerability in the Apache ActiveMQ component with a CVSS score of 8.8. An authenticated attacker can abuse the Jolokia JMX-HTTP bridge exposed by ActiveMQ to trigger remote code execution on the broker's JVM. Hitachi Energy has released the EP2 patch, which upgrades ActiveMQ to version 5.19.5 and updates related OpenJDK and client components.
Google has launched EmbeddingGemma 2, a Gemma 4-based embedding model released under Apache 2.0 that places text, code, images, video and audio in one shared 768-dimensional vector space. It's modular, so developers load only the encoders they need: 270M parameters for text and code alone, scaling up to 740M for every modality. Google reports a 14% jump over EmbeddingGemma 1 on the MTEB Code benchmark while keeping the earlier model's multilingual text accuracy.
Google Research released results from five partner-led evaluations of its Population Dynamics Foundation Model (PDFM), a Google Earth AI system that turns search trends, mobility, built-environment density and environmental data into monthly location embeddings. Health institutions plugged these pre-trained embeddings directly into their own existing statistical and ML pipelines, without any task-specific fine-tuning, and tested them across five very different epidemiological problems. Each partner reported measurable gains over conventional data sources, arguing for PDFM as a plug-and-play substitute or supplement to slow, siloed health datasets.
Z.ai's GLM 5.3, a 753-billion-parameter mixture-of-experts model optimized for coding and long-horizon agentic tasks, is now available as a fully managed model on Amazon Bedrock. It builds on GLM 5 with improved coding benchmark results, standout cybersecurity capabilities (CyberGym score of 84.5), and deeper Bedrock integration including OpenAI-compatible APIs, prompt caching, cross-Region inference and service tiers. AWS also demonstrates using GLM 5.3 with the open-source Strix agent for authorized penetration testing of your own applications.
Google Research released a report from its Contextual Agent Privacy and Security (CAPS) workshop, co-authored with over 50 academic and industry researchers. The report argues that AI agents need to reason about context-specific social norms, not just static permissions, to act appropriately when handling personal data and taking actions on a user's behalf. It proposes a multi-layered approach spanning system-level sandboxing, model-level reasoning, user-centric controls, multi-agent guardrails, and ecosystem governance, grounded in the theory of Contextual Integrity.
GitHub has published ReviewBench, a research-preview benchmark designed to compare AI code review agents on a common, auditable standard. The dataset draws on 219 real public pull requests across 19 languages and 187 repositories, chosen to mirror the language and size distribution found across 103.9 million GitHub pull requests. Ground truth comes from a blended pool of human reviewers, static analysis tools, and multiple frontier LLMs, cross-checked by senior engineers who agreed with the benchmark's labels 96.6% of the time. GitHub says it has already used ReviewBench internally to track its own Copilot code review (CCR) product, and that offline benchmark movement reliably predicted the direction of later production A/B tests.
Google has begun a global rollout of 'skills' inside the Gemini chat app, a feature that lets the assistant turn recurring chat instructions into reusable shortcuts it can trigger automatically on matching prompts. Users can combine several skills at once to handle more involved tasks, such as layering a tone-of-voice preference together with company-specific formatting rules. Google plans for skills to eventually take over the role currently played by Gems, and existing Gems will be converted into skills automatically once that tool is phased out. For now the feature is live only for users 18 and older, with wider availability to follow.
CISA added two Zammad vulnerabilities to its Known Exploited Vulnerabilities catalog after confirming active exploitation. CVE-2026-102489 is a session fixation flaw and CVE-2026-102490 involves improper privilege management in the Zammad helpdesk/customer service platform. Federal agencies must remediate under Binding Operational Directive 26-04, and CISA urges all organizations to prioritize these fixes.
AWS researchers describe a system that models telecom networks as a continuously updated graph ('digital twin') and uses AI agents to run a cascade of graph algorithms (decomposition, clustering, centrality ranking) to pinpoint root causes of failures. The approach was demonstrated with NTT DOCOMO at Mobile World Conference, reportedly cutting root cause analysis from hours to minutes on commercial networks. Agents also consult an incident knowledge base and runbooks, and can escalate autonomy from advisory to supervised to full remediation over time.
CISA added CVE-2026-104286, a path traversal vulnerability in Fortinet FortiMail, to its Known Exploited Vulnerabilities (KEV) Catalog after confirming active exploitation in the wild. Under Binding Operational Directive 26-04, federal civilian agencies must prioritize patching this vulnerability on internet-exposed assets, especially where it could grant full control of the system. CISA also recommends all organizations, not just federal agencies, prioritize remediation of KEV-listed flaws.
GitHub Security Lab built an open-source 'Taskflow Agent' that uses custom AI prompt workflows to audit code for vulnerabilities, and used it to find and report over 24 vulnerabilities in Android applications. Two detailed examples show the agent finding a location-tracking flaw in the OsmAnd navigation app and an account-takeover chain in the Wikipedia Android app via a deeplink parsing bug. The team notes LLMs are strong at finding complex logic vulnerabilities and understand API behavior well, but struggle to correctly judge severity and often produce false positives, so findings still require human review. The taskflows are open source and can be run on any GitHub repository via a Copilot-powered codespace script, though running them consumes premium model requests and can take hours.
Cisco Talos disclosed a China-nexus threat cluster, tracked as UAT-11587, that has targeted government, defense, diplomatic and policy organizations across Taiwan, India, the Philippines, Cambodia, Pakistan, Thailand, Myanmar and Syria since September 2025. The group delivers a previously undocumented Rust-compiled Windows backdoor called Antino via spear-phishing with spoofed senders and fake Gmail attachment widgets. Antino communicates covertly through Microsoft 365 (Outlook/OneDrive via Microsoft Graph) instead of a traditional C2 server, and the group heavily relies on Cloudflare infrastructure for delivery and staging. Talos found roughly 350 compromised endpoints across eight countries, with the largest wave hitting around 57 Indian endpoints in a single two-day period.
CERT-EU has updated its advisory on critical vulnerabilities in Microsoft SharePoint Server, confirming active exploitation of CVE-2026-50522 (CVSS 9.8), a deserialisation flaw allowing remote code execution. This is part of a wider series of recently patched critical bugs — CVE-2026-32201, CVE-2026-45659, CVE-2026-56164 and CVE-2026-58644 — affecting on-premise SharePoint Server instances, some exploitable without authentication.
Microsoft Threat Intelligence documented exploitation of CVE-2026-73570, an unauthenticated OS command-injection flaw in the Zimbra Collaboration Suite's SNMP notification path. Attackers triggered it via specially crafted emails against internet-facing servers running the optional zimbra-snmp package with SNMP notifications enabled, requiring no authentication or user interaction. Post-exploitation activity included JSP webshell deployment, privilege escalation to root, persistent access, credential and mailbox theft, lateral movement across Zimbra clusters, and custom remote-access malware. Zimbra fixed the issue in version 10.1.20 (released July 20, 2026); the CVE was publicly disclosed August 13, 2026, and Microsoft observed scanning and exploitation activity during the gap between patch availability and disclosure, as well as after.
On 9 September 2026, Check Point released emergency hotfixes for two critical vulnerabilities (CVE-2026-85102 and CVE-2026-85103), both with CVSS 9.8, affecting Security Gateway, Security Management Server, and Spark Firewall deployments using Remote Access VPN or Site-to-Site VPN. Both flaws allow unauthenticated remote attackers to execute arbitrary code on affected appliances. CERT-EU urges immediate patching, prioritising internet-facing and perimeter devices.
Citrix disclosed 8 vulnerabilities affecting customer-managed NetScaler ADC and NetScaler Gateway, including 2 critical unauthenticated RCE flaws with CVSS 9.5 that Citrix confirms are being actively exploited. CERT-EU urges immediate patching and compromise assessment for internet-facing appliances. Other flaws include HTTP request smuggling, memory overflow bugs leading to RCE/DoS, a policy bypass, and a TCP ISN prediction issue.
Anthropic has launched the Claude Frontier Academy, a $100 million program designed to address a shortage of skilled staff capable of implementing AI inside large organizations. The goal is to certify a large number of so-called Frontier Deployed Engineers by the close of 2027, with early participants coming from firms including Accenture, Bain, Capgemini, Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley and Novo Nordisk. The flagship course, a residency modeled loosely on medical training, combines an in-person intensive with a 12-week supervised project inside the engineer's own company before certification.
AWS released a step-by-step walkthrough showing how to give Claude Desktop (on Amazon Bedrock) live internet access through Amazon Bedrock AgentCore Gateway, closing the gap left by the model's training cutoff. The setup wires AWS IAM Identity Center SSO through Amazon Cognito as a SAML-to-OAuth federation layer, issuing JWTs that the AgentCore Gateway validates on every request. The Web Search capability itself is MCP-compatible and runs on an internal Amazon web index covering tens of billions of documents, so no query traffic leaves AWS and no third-party API key is needed. The feature is currently limited to three AWS Regions: us-east-1 (N. Virginia), eu-west-1 (Ireland), and ap-northeast-1 (Tokyo).
NVIDIA released a detailed workflow for fine-tuning its Nemotron 3.5 ASR multilingual streaming model on Saudi Arabic dialects (Najdi and Hijazi). Using minimal data curation, replay mixing with FLEURS data, duration-based bucketing, and partial encoder unfreezing, they reduced word error rate on the target dialects from 55.05% to 29.96% while also slightly improving English and other Arabic performance. The post also covers decoding tweaks and speaker diarization extensions for multi-speaker transcription.
Allen Institute for AI (Ai2) has open-sourced AstaBrief 8B, a model built on Qwen3-8B that turns research questions and retrieved literature excerpts into cited scientific reports. It's now live in Ai2's Asta platform as a 'Fast mode' alongside a Claude-powered 'Thinking mode', generating reports in about 51.1 seconds on average compared to 178.5 seconds for the Claude pipeline — roughly 3.5x faster. Ai2 trained the model using supervised fine-tuning on 47K examples (from 90K filtered real research queries) followed by DPO on about 6K preference pairs, and found that filtering training data for citation density was the single most effective lever for improving grounding quality.
Microsoft Research has unveiled Quine, a research system combining a multimodal 'world model' of biology with an interactive harness linking AI models, scientific literature, lab tools, and researchers. Built jointly across data types such as genomics, proteins, chemistry, RNA/cell state, and bioimaging, the model aims to predict how biological systems respond to interventions before costly lab experiments are run. Working with the Broad Institute of MIT and Harvard, Microsoft used Quine to rank thousands of compounds for their potential to shift pancreatic cancer cells between therapeutic states, and the top-ranked candidates were validated in wet-lab assays within a single weekend. Microsoft is now opening a 'Quine Fellows' program to give selected scientists early access, while cautioning that the system is experimental, research-only, and not intended for clinical use.
A Google Cloud TPU engineering team worked with Ai2 to rebuild Olmo 3 7B's pre-training from scratch using MaxText, Google's JAX/XLA training framework, running on TPUs instead of Ai2's original PyTorch/GPU setup. They matched Ai2's published results not just on the training loss curve but on four independent held-out evaluation surfaces across the full ~5.93-trillion-token, 1.41-million-step stage-1 run plus the stage-2 annealing phase. Along the way they ported Olmo 3's unusual architecture (reordered-norm blocks, QK-norm, 3:1 sliding/global attention ratio) into MaxText and caught a data-loader bug that had been quietly inflating apparent performance through memorization rather than genuine learning.
Google engineers describe how they implemented sparse spatio-temporal attention for video diffusion models on TPU v6e chips, converting the theoretical sparsity of the Sparse VideoGen (SVG) approach into real hardware speedups. Through a progression of kernel optimizations (full/boundary tile specialization, tile-size tuning, and mask-tile alignment), they reduced attention kernel latency from 96.37ms (naive sparse) to 32.76ms, a 2.40x speedup over dense Splash Attention on a single TPU v6e chip, using 75.6K tokens, 10 heads, and head dimension 128.