Only 9% of teachers have received formal AI policies

Teachers are struggling to navigate the integration of artificial intelligence in the classroom due to a lack of official guidance. A recent Gallup survey found that only 9% of teachers have received formal policies for specific AI-related tasks, while 48% have received informal guidance. Experts stress the need for comprehensive policies to support teachers in adopting AI.

New Mexico lawmakers are considering adopting policies to define AI use in schools, addressing issues like data privacy, curriculum, and tribal sovereignty. This move comes as schools and districts are looking to provide more comprehensive guidance on AI adoption.

Researchers have developed LLMSurgeon, a framework for post-hoc analysis of large language model pretraining data mixtures. This development addresses the need for auditing foundation models and understanding model behavior, providing a verifiable benchmark for transparency in LLM auditing.

Companies are also working on enhancing AI security, with Xage Security announcing enhanced Zero Trust capabilities for AI agents. This provides a secure production environment and comprehensive visibility and control over AI agents, mitigating risks associated with unauthorized actions and data exfiltration.

AI is being explored for its potential to predict and manage drought, with a new book discussing the benefits and challenges of using AI in drought prediction and management. AI tools can analyze historical climate records, current observations, and satellite imagery to assess and forecast drought.

In the payments industry, companies like Western Union and Remitly are leveraging AI to improve efficiencies and streamline operations. The industry is shifting towards AI-driven solutions to improve customer experiences and expand services.

New tools are emerging to help manage AI agents, including Salus, a runtime guardrails proxy that validates AI agent actions before they execute. This provides an additional layer of security and control for AI systems.

Experts are also discussing the differences between traditional observability and agent observability, highlighting the challenges of evaluating AI agent performance and the need for specialized tools and approaches.

A recent audit found that at least 146,932 citations in scientific papers published in 2025 are hallucinated, or fabricated, by large language models. This issue highlights the risks of contamination and bias in scientific recognition.

The AI boom is driving a resurgence of interest in hardware, with the semiconductor industry experiencing a revival and new investments and innovations emerging.

Key Takeaways

• Only 9% of teachers have received formal policies for specific AI-related tasks, while 48% have received informal guidance. • New Mexico lawmakers are considering policies to define AI use in schools, addressing issues like data privacy and curriculum. • LLMSurgeon is a new framework for post-hoc analysis of large language model pretraining data mixtures. • Xage Security has announced enhanced Zero Trust capabilities for AI agents, providing a secure production environment. • AI is being explored for its potential to predict and manage drought. • Companies like Western Union and Remitly are leveraging AI to improve efficiencies and streamline operations in cross-border payments. • Salus is a runtime guardrails proxy that validates AI agent actions before they execute. • At least 146,932 citations in scientific papers published in 2025 are hallucinated, or fabricated, by large language models. • The AI boom is driving a resurgence of interest in hardware, with the semiconductor industry experiencing a revival.

Most Teachers Lack Official AI Guidance

A recent Gallup survey found that few teachers have received official guidance on using artificial intelligence in the classroom. Only 9% of teachers reported getting formal policies for specific AI-related tasks. About 48% received informal guidance, while a third of teachers got no guidance at all. Experts say schools should provide more comprehensive policies to help teachers navigate AI adoption.

New Mexico Pushes for AI Education Policies

New Mexico lawmakers are being urged to adopt policies defining the use of artificial intelligence in schools. A legislative analyst recommends addressing issues like data privacy, curriculum, and tribal sovereignty. The state's current guidance for AI use in classrooms is non-binding, and lawmakers are considering comprehensive policies to support teachers and students.

Unlocking LLM 'Digital DNA' with LLMSurgeon

Researchers have developed a new framework called LLMSurgeon, which enables post-hoc analysis of large language model pretraining data mixtures. This breakthrough addresses the need for auditing foundation models and understanding model behavior. LLMSurgeon provides a verifiable benchmark for transparency in LLM auditing.

Strengthening AI Security with Zero Trust

Xage Security has announced enhanced Zero Trust capabilities for AI agents, providing a secure production environment and comprehensive visibility and control over AI agents. This development addresses critical needs in Cloud AI security, mitigating risks associated with unauthorized actions and data exfiltration.

Can AI Help Predict and Manage Drought?

A new book explores the potential of AI in predicting and managing drought. AI tools can analyze historical climate records, current observations, and satellite imagery to assess and forecast drought. Experts discuss the benefits and challenges of using AI in drought prediction and management.

Cross-Border Payments Embrace AI

Companies in cross-border payments are leveraging AI to improve efficiencies and streamline operations. Western Union and Remitly are using AI to enhance productivity and reduce costs. The industry is shifting towards AI-driven solutions to improve customer experiences and expand services.

Salus: The Bouncer for AI Agents

Salus is a runtime guardrails proxy that validates AI agent actions before they execute. This tool helps prevent AI agents from confidently doing the wrong thing at scale. Salus provides an additional layer of security and control for AI systems.

Agent vs. Traditional Observability

Phil Hetzel, Head of Solution Engineering at Braintrust, explains the differences between traditional observability and agent observability. He highlights the challenges of evaluating AI agent performance and the need for specialized tools and approaches.

AI Hallucinations in Scientific Papers

A recent audit found that at least 146,932 citations in scientific papers published in 2025 are hallucinated, or fabricated, by large language models. The issue of AI hallucinations in research papers is growing, and experts warn about the risks of contamination and bias in scientific recognition.

Hardware is Sexy Again

Plug and Play CEO Saeed Amidi discusses the resurgence of interest in hardware, driven by the AI boom. The semiconductor industry is experiencing a revival, with new investments and innovations emerging.