Archive

All Posts

OpenAI Restricts Astra AI Model Over Potential Critical Cyber Risk

AI Safety and Security

Aug 10, 2026

OpenAI Restricts Astra AI Model Over Potential Critical Cyber Risk

OpenAI has not determined whether Astra is Critical, but that possibility alone has already changed how the model can be developed.

Alicia Shapiro

Meta Launches Muse Code for Long AI Coding but Keeps Its Model Closed

New AI Products and Features

Aug 7, 2026

Meta Launches Muse Code for Long AI Coding but Keeps Its Model Closed

Would your engineering team accept a closed AI coding system to cut costs, or does long-term control matter more?

Alicia Shapiro

Anthropic & AISI: Agents Reached Real Systems, Exposing Security Gaps

Agentic Workflows

Aug 5, 2026

Anthropic & AISI: Agents Reached Real Systems, Exposing Security Gaps

Anthropic’s and AISI’s evaluation reports reveal a hard problem: tests designed to expose dangerous AI behavior can also give agents room to reach real targets.

Alicia Shapiro

LinkedIn Expands AI Slop Reporting Without Clear Creator Safeguards

AI Governance

Jul 31, 2026

LinkedIn Expands AI Slop Reporting Without Clear Creator Safeguards

If readers cannot reliably identify AI writing, an AI-slop report can become a judgment about whose work belongs in the professional conversation.

Alicia Shapiro

AI Workers Ask U.S. Government for Tools to Slow AI Before a Crisis

AI Research

Jul 29, 2026

AI Workers Ask U.S. Government for Tools to Slow AI Before a Crisis

If governments wait until frontier AI is moving too fast to manage, building an international slowdown may already be impossible.

Alicia Shapiro

Claude Opus 5 Launch Aims to Finish AI Work With Fewer Human Fixes

Agentic Workflows

Jul 27, 2026

Claude Opus 5 Launch Aims to Finish AI Work With Fewer Human Fixes

Anthropic wants Claude Opus 5 to take over AI’s last mile: turning a promising answer into dependable finished work.

Alicia Shapiro

OpenAI Presence Gives Enterprises Control Over Voice and Chat Agents

AI Voice

Jul 24, 2026

OpenAI Presence Gives Enterprises Control Over Voice and Chat Agents

What happens when a voice or chat agent’s correct answer or approved action becomes outdated after launch?

Alicia Shapiro

OpenAI Models Bypassed a Cyber Test, Changing AI Safety Evaluations

AI Research

Jul 22, 2026

OpenAI Models Bypassed a Cyber Test, Changing AI Safety Evaluations

A benchmark can record a correct answer, even when the model found a way around the test to get it.

Alicia Shapiro

OpenAI’s New Scorecard Measures AI’s Business Value by Completed Work

Reasoning Models

Jul 20, 2026

OpenAI’s New Scorecard Measures AI’s Business Value by Completed Work

A low token price can become expensive when employees still have to review, correct, retry, rework, or finish the work AI was meant to complete.

Alicia Shapiro

Archive

OpenAI Restricts Astra AI Model Over Potential Critical Cyber Risk

Meta Launches Muse Code for Long AI Coding but Keeps Its Model Closed

Anthropic & AISI: Agents Reached Real Systems, Exposing Security Gaps

LinkedIn Expands AI Slop Reporting Without Clear Creator Safeguards

AI Workers Ask U.S. Government for Tools to Slow AI Before a Crisis

Claude Opus 5 Launch Aims to Finish AI Work With Fewer Human Fixes

OpenAI Presence Gives Enterprises Control Over Voice and Chat Agents

OpenAI Models Bypassed a Cyber Test, Changing AI Safety Evaluations

OpenAI’s New Scorecard Measures AI’s Business Value by Completed Work

AiNews.com