snowflake unveils new tools as google ships new models

Key Takeaways

• Snowflake is launching an iOS app and integrating with Salesforce, Gmail, and Slack to act as a personalized work agent that executes tasks directly.
• Snowflake Cortex Code AI now supports Databricks and AWS Glue, enabling engineers to manage cross-system data and migrate Spark workloads.
• Snowflake's Cortex Agents platform includes native Model Context Protocol support and a sandboxed Python environment for secure code execution.
• Accuity won the 2026 AI Excellence Award, having processed over 7 million inpatient charts and delivered $3.3 billion in revenue for healthcare clients.
• A study shows LinkedIn is cited 14.3% of the time in ChatGPT Search responses, surpassing Wikipedia and major news publishers for credibility.
• Researchers found AI tools in conflict zones score 0.08/1.0 on Women Peace and Security standards unless explicitly prompted to include them.
• San Diego State University uses AI to analyze aerial imagery and create a real-time map of homelessness to optimize outreach efforts.
• Cisco released a security scanner to protect AI agents in IDEs like Cursor and VS Code from vulnerabilities and unauthorized file access.
• Prisoners at Jessup Correctional Institution are accessing ChatGPT via friends to research legal arguments regarding inmate health screenings.
• Electronic music pioneer Jean-Michel Jarre urges the film and music industries to embrace AI as an augmented imagination tool rather than fearing it.

Snowflake Intelligence becomes a personalized work agent

Snowflake is transforming its platform into a personalized work agent that helps users take direct action across business tools. This new feature allows employees to ask questions about enterprise data and immediately perform tasks without switching applications. Integrations with tools like Gmail, Google Calendar, Jira, Salesforce, and Slack enable seamless workflows through upcoming MCP connectors. An iOS mobile app is also entering public preview to extend this capability. The system learns how individuals access data and interact with their tools to eliminate the daily grind of chasing analysts or waiting for reports.

Snowflake Cortex Code AI supports more data systems

Snowflake's Cortex Code AI assistant now understands and integrates across a broader data stack including AWS Glue, Databricks, and Postgres. This expansion allows data engineers to maintain context across disparate systems and reduces time spent debugging cross-system failures. The platform introduces specialized agent skills for core data engineering tasks such as Snowpark Python, data quality checks, and cost intelligence. New skills like snowpark-connect help migrate Apache Spark workloads to Snowflake while dbt-projects-on-snowflake manages native Snowflake object management for dbt projects. Teams can now program the agent using Python and TypeScript to embed its capabilities into their own tools and workflows.

Snowflake launches Cortex Agents for enterprise AI development

Snowflake has formalized its Cortex Agents platform to streamline enterprise AI agent development with native support for the Model Context Protocol. This integration connects AI agents to services like Jira, Salesforce, and Slack while reducing custom engineering burdens. The platform includes a Code Execution Tool that provides a sandboxed Python environment for on-demand code generation and execution. Agent Skills are modular, reusable packages that allow organizations to codify domain expertise into shareable components. Multi-tenancy support ensures data isolation between user groups using row access policies and immutable session attributes.

Enterprise AI success depends on organizational architecture

Sanjay Srivastava argues that successful enterprise AI adoption requires rethinking organizational architecture rather than just focusing on models or tools. Companies must define clear boundaries between centralized innovation and federated business-level execution to balance focus with energy. The AI layer architecture involves managing optionality and understanding day two costs beyond simple inference expenses. Underlying enterprise technology stacks must also evolve as AI becomes the interface layer for orchestration and control planes. Durable advantage comes from making coherent choices across organizational structure, AI layer design, and technology stack logic.

Prisoners seek legal help using AI chatbots

Jason Allen, a prisoner at Jessup Correctional Institution, used an outside friend to access ChatGPT for potential legal arguments regarding inmate cancer screenings. Although inmates cannot freely browse the internet or use AI directly, they can request friends to send screenshots of chatbot answers through the prison messaging system. Many inmates are eager to learn about AI by attending workshops or asking friends to send printouts by snail mail. Some prisoners even use contraband cellphones to gain access to the technology despite strict restrictions on internet and AI tools.

Jean-Michel Jarre urges music and film industries to embrace AI

Electronic music pioneer Jean-Michel Jarre criticized the conservatism of the music and film industries for fearing AI instead of embracing it. He stated that artists should use AI to create the cinema and music of tomorrow, comparing it to the introduction of electricity or the violin. Jarre, who has sold 85 million records and is a former president of the International Confederation of Societies of Authors and Composers, spoke at the second AI film festival in Cannes. He argued that AI is not a threat but an augmented imagination tool that will help create new genres of cinema and music.

San Diego State University uses AI to map homelessness

Researchers at San Diego State University are building a comprehensive real-time map of homelessness using AI to analyze aerial photos and street view images. The interactive map shows hot spots of homelessness in specific neighborhoods plus movement and migration patterns throughout the year. This data helps predict where to send outreach workers and street health teams more effectively. The map also identifies shelters, food banks, health clinics, and other outreach services to help homeless people find safety and community resources.

Accuity wins 2026 AI Excellence Award for healthcare

Accuity was named a winner in the 2026 Artificial Intelligence Excellence Awards for advancing responsible AI in healthcare. The company uses its Amplifi AI engine to review inpatient charts and ensure clinical reality is accurately captured in data used for reimbursement. Physicians independently interpret signals from the AI and apply accurate clinical coding outcomes within client systems. This approach has reviewed more than 7 million inpatient charts and delivered $3.3 billion in incremental revenue to clients. Accuity supports more than 65 health systems and 450 hospitals nationwide with a team of more than 200 physicians.

Defense AI must follow Women Peace and Security frameworks

Researchers found that AI tools used in conflict zones systematically fail to operationalize Women Peace and Security standards unless explicitly prompted. Testing three leading AI models across 13 conflict scenarios showed performance drops when WPS language was omitted from prompts. When prompts explicitly named affected populations like displaced women, average scores were 0.65 out of 1.0, but minimal formats dropped scores to 0.08. This gap indicates that deployed AI tools do not factor in women in their analysis, which can have real consequences for over 50 percent of the global population.

AI search engines prioritize LinkedIn for credibility

A new study analyzed 325,000 unique prompts across major AI search platforms and found that LinkedIn appeared in 11 percent of AI-generated responses on average. On ChatGPT Search specifically, LinkedIn citations climbed to 14.3 percent, ahead of Wikipedia, YouTube, and major news publishers. The study measured semantic similarity and found that AI responses closely mirrored the meaning of LinkedIn content with scores between 0.57 and 0.60. This means every LinkedIn article or thought leadership post actively shapes how topics are explained to users who may never visit the original page.

New security scanner protects AI agents in IDEs

Cisco introduced an AI Agent Security Scanner for IDEs to verify the security of AI agents in tools like Cursor, VS Code, and Windsurf. These agents utilize Model Context Protocol servers and can access file systems, APIs, and shell commands, creating new attack surfaces. The scanner operates on a defense-in-depth model with proactive vulnerability prevention during code generation and static analysis of server configurations. It includes capabilities for MCP Server Scanning, Agent Skill Scanning, secure AI-generated code verification, and continuous Watchdog monitoring. The tool integrates natively into the IDE experience with a security dashboard and inline decorations for findings.

Sources

• Snowflake's Agent Turns Data Into Action
• Snowflake's AI Coder Goes Full Stack
• Snowflake's Cortex Agents Go Live
• Architecture: A Question At The Core Of AI In The Enterprise
• Even Without Internet Access, Prisoners Are Trying to Benefit From A.I.
• Jean-Michel Jarre urges music and film industries to embrace AI
• Using AI to map homelessness
• Accuity Named Winner in 2026 Artificial Intelligence Excellence Awards for Advancing Responsible AI in Healthcare
• Women, Peace and Security Frameworks Must Apply to Defense AI
• AI Is Studying Your LinkedIn Profile—And What It Finds May Decide Your Credibility
• Introducing the AI Agent Security Scanner for IDEs: Verify Your Agents