Meta releases Omnilingual ASR to expand global speech recognition 

Meta has introduced Omnilingual ASR, an open-source automatic speech recognition suite that supports more than 1,600 languages at launch. The system uses zero-shot in-context learning, allowing developers to extend coverage to over 5,400 languages by providing only a few example audio and text pairs, with no retraining required. The release includes self-supervised audio encoders, CTC-based ASR models, large-scale LLM-based decoders, and a multilingual audio model with 7 billion parameters. Meta is also releasing an extensive speech dataset that includes many underrepresented languages, all provided under the Apache 2.0 license. This development provides enterprises with new opportunities to build voice interfaces, transcription tools, and accessibility features across regions and languages without restrictive licensing costs.

Google launches Workspace Studio to help users automate everyday work

Google has introduced Workspace Studio, a new platform that allows users to build custom AI agents that automate routine tasks across Gmail, Docs, Sheets, Drive, Meet, and other Workspace applications. Users can describe what they want in plain language, and the system automatically generates agents that can handle email sorting, document processing, scheduling, report creation, and multi-step workflows. The platform also supports collaboration, allowing teams to share and adapt agents based on their needs. Workspace Studio connects with third-party tools such as Asana, Jira, Mailchimp, and Salesforce, which expands its automation capabilities beyond core Workspace functions.

NetSuite highlights how AI is transforming ERP capabilities

NetSuite is transforming ERP with AI by automating processes, delivering actionable insights, and enabling more intuitive, conversational user interactions. AI capabilities are embedded across multiple modules, allowing companies to leverage advanced analytics, predictive insights, and automation features to simplify tasks such as financial reconciliations, customer-service inquiries, and planning workflows. Recent releases introduced three AI-powered tools that enhance productivity: generative text creation, automated invoice data capture to eliminate manual entry, and intelligent data analysis for planning and budgeting. Financial exception management is also now available, providing teams with faster, more accurate ways to identify and resolve anomalies.

SAP’s AI powered support strategy delivers proactive, instant customer service

SAP is transforming its customer support model by embedding AI across support operations to anticipate issues, automate responses, and deliver instant resolutions at scale. With its AI support ecosystem, SAP can predict potential problems such as system overloads or platform performance issues before they impact users, helping ensure uninterrupted operation even during peak demand periods. Over 80 percent of customer issues are now resolved via self‑service tools and AI agents. This shift does not replace human specialists. Rather it frees support engineers from routine tasks so they can focus on complex issues requiring expert judgement. For companies using SAP, this means faster support, reduced downtime risk, and smoother scaling of AI or cloud deployments.

Fake ChatGPT and Grok ads distribute macOS info stealer malware

Security researchers have uncovered a malicious campaign that uses paid Google search ads to promote shared ChatGPT and Grok conversation‑guides which secretly deliver the AMOS Infostealer to macOS users. The campaign begins when users search for common Mac troubleshooting terms. The ads link to seemingly legitimate AI‑chat guides hosted on official LLM platform domains. Once users follow the instructions and execute a shell command in the terminal, the AMOS malware is downloaded and installed with root privileges. AMOS is a full‑featured infostealer that targets browser cookies and passwords, cryptocurrency wallet apps, Keychain credentials, and stored files. It also installs a persistent backdoor for long‑term access. This attack shows how threat actors exploit trust in AI tools and legitimate ad platforms to distribute malware at scale.

OWASP releases top 10 risks and mitigations for agentic AI security

The OWASP GenAI Security Project has published the first “Top 10 for Agentic Applications,” offering a vendor‑neutral, community‑backed framework to help organizations identify and mitigate major security risks when deploying autonomous AI agents. Among the top threats are goal hijacking (agents redirected to malicious objectives), tool misuse or exploitation, identity and privilege abuse, supply‑chain vulnerabilities, and rogue agents acting beyond their intended scope. The release comes with practical mitigation guidance covering secure prompt handling, privilege management, supply‑chain hygiene, runtime monitoring, and robust governance mechanisms to manage agent behavior across development, deployment, and operations.