Google’s Gemini AI revolutionizes scientific research with automated empirical software 

Google has introduced an advanced AI system powered by Gemini and designed to automate the creation of empirical software, specifically across fields such as bioinformatics, epidemiology, and neuroscience. By leveraging large language models (LLMs) and tree search algorithms, the AI can generate, evaluate, and refine software tailored to specific research tasks. In epidemiology, it has produced more accurate COVID-19 forecasting models, while in neuroscience it has advanced whole-brain activity prediction and hybrid neuron simulations. The system frames research problems as “scorable tasks,” enabling iterative improvement without extensive human intervention.

Anthropic enhances Claude AI with persistent memory and chat search features

Anthropic has introduced a new memory feature for its Claude AI chatbot, enhancing its ability to support ongoing projects and maintain context across interactions. This feature is currently available to Team and Enterprise users, allowing Claude to remember user preferences, project details, and priorities without the need for prompts. The memory function is optional and can be managed through settings, including the ability to disable it entirely or use an incognito mode for private conversations. Additionally, Claude now offers chat search capabilities, enabling users to reference past conversations for continuity. These updates aim to improve user experience by providing more personalized and efficient interactions with the AI.

NetSuite introduces AI-powered translation in Text Enhance to support 22 languages 

NetSuite has introduced Text Enhance Translate, a new addition to its generative AI–powered Text Enhance suite, enabling users to instantly translate any text field in NetSuite into one of 22 supported languages, with automatic detection of the source language. Unlike basic translation tools, this feature is context-aware, drawing on business workflows, operational data, tone, and system-specific terminology to deliver more accurate and relevant results. Administrators can configure which target languages are available, control access for external roles, and even extend translation into custom workflows through Prompt Studio or the SuiteScript GenAI API, making it a flexible solution for global operations.

SAP expands Sovereign Cloud to accelerate regional AI innovation in Europe

SAP has expanded its Sovereign Cloud offerings to bolster Europe's digital sovereignty and accelerate regional AI innovation. The enhanced portfolio provides European customers with a comprehensive technology stack, including SAP Cloud Infrastructure and SAP Sovereign Cloud On-Site, ensuring secure, compliant, and scalable innovation. This expansion allows organizations to choose deployment models tailored to their specific regulatory and operational needs, enabling them to innovate securely and in compliance with local laws while maintaining the freedom to scale on their own terms. With SAP Sovereign Cloud, customers can run their SAP Business Suite in sovereign environments while benefiting from continuous innovation cycles, including SAP Business Technology Platform (SAP BTP) and embedded SAP Business AI capabilities.

Workday confirms data breach amid wider Salesforce-linked attacks

Workday, the global HR software giant, has disclosed a data breach linked to the recent wave of attacks targeting Salesforce CRM systems, exposing limited business contact details such as names, emails, and phone numbers while confirming no customer tenants were affected. The breach, discovered on August 6, 2025, stemmed from a social engineering campaign in which attackers impersonated HR or IT staff to trick employees into granting access. This breach appears to be part of a larger campaign linked to the ShinyHunters extortion group. They have been targeting Salesforce CRM systems via social engineering and phishing (including voice phishing), with the aim of getting victims to install or authorize malicious OAuth apps. Once inside, attackers exfiltrate data and often follow up with extortion demands.

AI browsers vulnerable to “PromptFix” exploit in hidden prompts

Security experts at Guardio Labs have uncovered a new exploit called PromptFix, which allows attackers to trick AI-powered browsers into executing hidden malicious commands by embedding prompts in fake CAPTCHA checks or invisible web elements. This flaw, seen in agentic browsers like Perplexity’s Comet and even ChatGPT’s Agent Mode, can lead to unauthorized actions such as auto-purchasing items, filling saved payment details, or downloading malware without the user’s awareness. The attack combines social engineering with AI automation, exploiting the technology’s drive to complete tasks quickly and thoroughly. Researchers warn that stronger guardrails, including phishing detection, domain reputation checks, and strict oversight of AI-initiated actions, are essential to prevent misuse as AI browsing becomes more widespread.