Learn how we can help you drive your business

OpenAI introduces textGrain watermarks to identify AI-generated text
OpenAI is introducing textGrain, an invisible text watermark, to eligible ChatGPT and Codex outputs in the European Union over the coming weeks to meet the EU AI Act’s transparency requirements. Unlike conventional AI detectors that infer authorship from writing patterns, textGrain embeds a statistical signal through subtle adjustments to the model’s word choices. API customers worldwide can opt in for selected models, with watermarking disabled by default. OpenAI reports no meaningful impact on output quality, but detection remains vulnerable to short passages, translation and editing. The text detector will initially be available only to approved researchers and expert organizations. OpenAI stresses that a watermark cannot prove authorship, ownership or factual accuracy.
 
Shopify enables browser-based AI agents to complete checkout with user approval 
Shopify has expanded WebMCP support to checkout, enabling browser-based AI agents to move beyond product discovery and cart creation to completing purchases on eligible merchant storefronts. Three tools, get_checkout, update_checkout and complete_checkout, let agents retrieve checkout details, modify information such as delivery addresses and shipping options, and submit orders after buyer authorization. Built on the Universal Commerce Protocol (UCP), WebMCP provides structured commerce data and actions, reducing the need for agents to repeatedly inspect web pages and simulate clicks. Shopify also preserves merchant checkout rules and returns control to buyers when payment verification is required. In Shopify’s test using GPT-6 Sol, WebMCP reduced operation time from 27.4 to 10.3 seconds and testing costs by 58%.
 
NetSuite 2026.2 brings Ask Oracle conversational AI to users
NetSuite 2026.2 introduces Ask Oracle, a conversational AI assistant from NetSuite Next, into the existing NetSuite experience. Users can ask questions in natural language to analyze business data, generate reports, create visualizations, locate records and dashboards, and complete everyday tasks without navigating multiple menus. For deeper analysis, users can move from chat to a canvas workspace to build ad hoc dashboards, KPI scorecards, charts, tables and written summaries, then export the results as PDFs. Ask Oracle operates within existing user roles and permissions, while its “Show Explanation” feature helps users understand how responses are generated. Initially, the capability is available to customers in the United States and Canada, with additional countries planned.
 
SAP Pay embeds automated payment execution into SAP Cloud ERP
SAP has introduced SAP Pay, a payment service embedded directly into SAP Cloud ERP to simplify supplier and treasury payments. Powered by Tereina, a regulated payment-service provider established in April 2026 as a wholly owned SAP subsidiary, SAP Pay connects invoice approval, payment execution and reconciliation within a single ERP workflow. Once an invoice is approved, the service can automatically execute payment and reconcile it against the corresponding invoice or purchase order, reducing manual bank file handling, reliance on external payment platforms and reconciliation delays. Supported methods include electronic funds transfer (EFT), including ACH, wire transfers and checks, as well as stablecoin-based settlement for cross-border transactions. SAP Pay is generally available in the United States and United Kingdom.
 
Anthropic expands cyber verification program to empower security teams
Anthropic is expanding its Cyber Verification Program (CVP), integrating Project Glasswing into a three-tier access framework that gives verified cybersecurity professionals access to advanced Claude models with fewer restrictions. Defense Access supports security operations, incident response, malware analysis and vulnerability research. Red Team Access adds authorized penetration testing and red-team activities, while Specialized Access is reserved for a limited number of verified organizations testing safety-critical systems such as power grids, aviation systems and interbank payment infrastructure. In testing, Claude Opus 5.5 completed 34 of 50 cyber-operation challenges under Red Team Access. Project Glasswing partners identified at least 129,000 verified software vulnerabilities between April and July 2026, with more than 33,000 rated critical or high severity.
 
Attackers abuse ChatGPT custom GPTs to spread remote access trojans
Researchers discovered a cyberattack campaign abusing ChatGPT Custom GPTs to impersonate legitimate services and redirect users to malicious websites. The attackers created GPTs named “Plus 5.6” and displayed fake service availability notices, directing users to a Google Sites page disguised as a Cloudflare CAPTCHA. This ClickFix technique tricked victims into manually running PowerShell commands that downloaded malicious installers and deployed a remote access trojan. The infection chain used obfuscated scripts, DLL sideloading through legitimate, digitally signed applications, and multiple persistence mechanisms to evade detection and maintain access. Huntress investigated at least 40 related incidents, including two confirmed infections originating from Custom GPTs, highlighting how trusted AI platforms can be exploited for social engineering.