Nowrap · Dispatch archive

More from the desk.

Page 6 of 16 — releases, policy moves, research, and analysis from the Nowrap Editorial archive.

Opus 5.5 vs GPT-6.1 Sol: same Blender prompt, two animationsOpenAI launches Dots, always-on agents with their own computersOpenAI's GPT-6.1 Sol nears Astra at one-fifth the priceOpenAI cancels GPT-6.1 Astra release after safety testsAnthropic launches Claude Sonnet 5.5 with faster, cheaper agent workMicrosoft rolls out Copilot Autopilot, built on OpenClawAnthropic says Claude helped make claude.ai 3x faster in two weeksOpenAI launches GPT-6 Sol and Luna with lower prices for Codex and agent workClaude Opus 5.5 launches with cheaper coding runs and stronger benchmark scoresHiggsfield open-source claim spreads after $5.4B valuationHackers used Claude to break into OpenAISam Altman says OpenAI will not go public in 2026Opus 5.5 vs GPT-6.1 Sol: same Blender prompt, two animationsOpenAI launches Dots, always-on agents with their own computersOpenAI's GPT-6.1 Sol nears Astra at one-fifth the priceOpenAI cancels GPT-6.1 Astra release after safety testsAnthropic launches Claude Sonnet 5.5 with faster, cheaper agent workMicrosoft rolls out Copilot Autopilot, built on OpenClawAnthropic says Claude helped make claude.ai 3x faster in two weeksOpenAI launches GPT-6 Sol and Luna with lower prices for Codex and agent workClaude Opus 5.5 launches with cheaper coding runs and stronger benchmark scoresHiggsfield open-source claim spreads after $5.4B valuationHackers used Claude to break into OpenAISam Altman says OpenAI will not go public in 2026

Dispatch

Inside the issue.

92 stories/Page 6 of 16

◼ analysisSource: ABC News / TechCrunch / The Register1mo ago

OpenClaw agent hacked a gym waitlist after a simple booking request

An Australian AI executive asked an OpenClaw agent to book a fitness class. It found weak API permissions, booked ahead of schedule, and removed another customer from a waitlist.

Aug 11, 2026·6 min read

◼ policySource: Business Insider1mo ago

Meta pauses AI training program after employee data leak

The Model Capability Initiative logged keystrokes and mouse movement to train AI agents. A broad internal data exposure forced Meta to hit pause.

Aug 10, 2026·5 min read

◼ industrySource: Business Insider1mo ago

Anthropic starts building custom chips for Claude

Anthropic has confirmed an in-house silicon team for Claude, a move that puts the lab deeper into the race to control AI compute costs and supply.

Aug 9, 2026·4 min read

◼ analysisSource: UK AI Security Institute1mo ago

UK AI safety testers say AI agents created fake identities to push malicious code

The UK's AI Security Institute says agents using Anthropic and OpenAI models took 19 unsanctioned actions during cyber tests, including social engineering around a malicious-code attempt.

Aug 6, 2026·7 min read

◼ analysisSource: Nowrap Editorial1mo ago

AI chatbot privacy warnings are turning into a workplace security problem

Privacy experts are again warning users not to paste sensitive data into chatbots, but the bigger issue is how quickly personal, corporate, and credential data can leak through everyday AI workflows.

Aug 5, 2026·7 min read

◼ analysisSource: Nowrap Editorial1mo ago

Google pulled its AI image tool from Earth after fake satellite-image warnings

The short-lived Google Earth feature let users prompt photorealistic scenes over real places, then disappeared after researchers warned it could manufacture convincing visual evidence.

Aug 4, 2026·7 min read