Top AI & Tech News Stories — August 5, 2026

OpenAI Astra AI Model May Reach Critical Cyber Capability Threshold

11 outlets · 14 articles · 70 total since July 21, 2026 · OpenAI, Anthropic, Hugging Face, United Kingdom, incident response

OpenAI has temporarily slowed frontier AI training after internal testing indicated that its upcoming Astra model could approach a critical cybersecurity capability threshold. The decision highlights growing concern that increasingly capable AI systems may independently…

More coverage: constellationr.com · the-decoder.com · cnbc.com · news.google.com · thehackernews.com

Matthew Green on Anthropic’s New Cryptanalysis Results

1 outlet · 1 article · cryptanalysis, Cryptography, Anthropic, Google, Matthew Green

Matthew Green: If you’re under the impression that these models are “glorified autocomplete” or that progress is slowing down, I need to urge you: stop thinking that. The models are very intelligent and capable, they are getting better at a fast clip. I can cite measurable and…

Meta Muse Code vs. Fable 5: Meta Muse is cheaper, but at what cost?

5 outlets · 6 articles · 12 total since August 5, 2026 · Meta, Anthropic, OpenAI, Muse Code, Muse Spark 1.2

Meta released Muse Code on August 5 , its first AI coding agent, built on the new Muse Spark 1.2 model. Mark Zuckerberg says it handles “complete software engineering tasks across large repos,” and that big jobs “fan out to separate sub-agents working in parallel.” Meta’s AI…

More coverage: thenewstack.io · techcrunch.com · cnbc.com · news.google.com

Paramount Skydance Caps Claude AI Tokens as Hollywood Giants Face Soaring Costs

4 outlets · 4 articles · 8 total since August 5, 2026 · microsoft, OpenAI, Jay Parikh, model selection, GitHub

David Ellison’s Paramount Skydance just drew a line. The entertainment giant has imposed monthly token limits on its tech employees’ use of Claude, Anthropic’s flagship AI model. Three Paramount employees told Business Insider the company wants to curb wasteful spending while…

More coverage: cnbc.com · news.google.com · theregister.com

Cloudflare OS is a Stripped-down Version of OpenClaw for Companies

1 outlet · 1 article · 2 total since August 5, 2026 · Open-source, Cloudflare, Cloudflare, AI agent platform, Autonomous apps

Cloudflare aims to move AI agents beyond the role of general-purpose chatbots and turn them into tools for everyday business work. With Cloudflare OS, the US company is introducing an open-source platform that allows employees to work with agents, use internal data, build their…

Google DeepMind resets leadership: Hassabis becomes chair, Kavukcuoglu takes operational lead - Dealroom

7 outlets · 11 articles · 45 total since August 5, 2026 · Google, deepmind, Demis Hassabis, Google DeepMind, Alphabet

Google DeepMind resets leadership: Hassabis becomes chair, Kavukcuoglu takes operational lead Dealroom

More coverage: nytimes.com · cnbc.com · news.google.com

Uber controls token spend, outlines AI use cases

1 outlet · 1 article · token spend optimization, open-weight models, AI inference, LLM, model selection

Uber controls token spend, outlines AI use cases Larry Dignan Wed, 5 Aug 2026 - 09:48 Larry Dignan Editor in Chief of Constellation Insights Constellation Research Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial…

Does Google even want to win at AI?

3 outlets · 3 articles · 9 total since August 5, 2026 · Google, Jeff Dean, Demis Hassabis, Discovery Loop, Google DeepMind

Today on Decoder , I’m talking with Hayden Field, The Verge ’s senior AI reporter, about a question that’s been rocketing around the tech industry for the past week: Is Google losing the AI race? That’s because last week Google announced a bombshell reorganization of its AI…

More coverage: thenewstack.io · news.google.com

Tom Cotton prods Treasury for tax code tweaks to modernize OT

1 outlet · 1 article · cybersecurity, industrial control systems, threat detection, Software Development, Tax compliance

With the country’s operational technology underfunded, outdated and increasingly vulnerable to cyberattacks, the Senate Intelligence Committee chairman is pushing the Treasury Department to do what it can to spur investment in it. Sen. Tom Cotton, R-Ark., said in a letter…

Perplexity Wins Appeal Against Amazon in AI Agent Shopping Lawsuit

2 outlets · 2 articles · AI agents, Perplexity, Amazon, agentic systems, web browsing

A U.S. appeals court ruled Amazon is unlikely to prove Perplexity violated federal hacking law, marking the first appellate decision on whether AI agents can legally act on users' behalf online.

More coverage: the-decoder.com

Justice league

1 outlet · 1 article

In this issue: White House decides not to disclose safety framework, but open models are exempt - For an undisclosed framework, its contents seem to be no great secret There Will Come Soft Rains - Today is August 4, 2026 Consumer protection laws for rogue AI - Nobody has the…

Zillow layoffs hit 91 jobs in Washington state, with senior roles bearing the brunt

1 outlet · 1 article

Zillow Group’s headquarters at 1301 Second Ave. in downtown Seattle. (GeekWire File Photo) Zillow Group’s layoffs will eliminate 91 jobs in Washington state, landing heavily on senior staff, according to a notice the company filed with the state Employment Security Department…

Google Earnings, The Frontier Case, Amazon Earnings

1 outlet · 1 article · Google, Anthropic, Amazon, Stratechery, Sinocism

Google’s earnings seemed to confirm the Anthropic hedge; it was Andy Jassy who explained why their — and Amazon’s — capex was justifiable. Subscribe to Stratechery Plus for full access. Already subscribed? $15 / month or $150 / year With Stratechery Plus you get access to the…

California to Ban AI Chatbots Marketed as Standalone Therapists in 2026

2 outlets · 2 articles · 19 total since July 16, 2026 · OpenAI, Anthropic, Hugging Face, California, sandboxing

California lawmakers have introduced Senate Bill 903, a measure that would prohibit the sale and use of artificial intelligence systems marketed as standalone therapists or mental health counselors starting in 2026. The legislation, which passed the Senate in May and now awaits…

More coverage: cyberscoop.com

Google will shut down Google Assistant starting September 2026 as Gemini takes over on Android and Wear OS

1 outlet · 1 article · Android, Wear OS, Android Auto, LLM, Google

Google is killing Google Assistant on Android and Wear OS starting September 4, 2026. Gemini takes over as the AI-powered successor on smartphones, tablets, watches, and in cars with Android Auto. Whether a probability-based LLM can match the reliability of its deterministic…

Africa’s AI Dividend Depends on Workflow Proof, Not Training Totals

1 outlet · 1 article · agent-assisted work, Data privacy, Accuracy measurement, repeatability, anomaly detection

Africa’s artificial intelligence debate is moving from possibility to delivery. A recent IMF analysis reported by Africa.com estimates that AI could expand sub-Saharan Africa’s economy by roughly 4 percent over the next decade if the region closes gaps in power, connectivity and…

South Africa Will Not Pick a Side Between US and Chinese AI Stacks – Malatsi

1 outlet · 1 article · Artificial intelligence, AI hardware, AI software, Open Standards, interoperability

South Africa will not align itself with either Washington or Beijing on artificial intelligence, communications minister Solly Malatsi has told TechCentral, rejecting the central premise of a new Boston Consulting Group study arguing that the window for countries to hedge…

Nairobi Securities Exchange Plans East Africa’s First AI ETF – But Its CEO Is Watching for a Bubble

1 outlet · 1 article · 17 total since July 17, 2026 · NVIDIA, Apple, SK hynix, South Korea, United States

The Nairobi Securities Exchange is building East Africa’s first exchange traded fund focused on AI stocks, with plans to offer it to investors before year-end — though its chief executive says the launch could be delayed if the global AI rally starts to look overdone. “We want…

He, it, but rarely she: When AI models write kids’ stories about animals, female characters vanish

1 outlet · 1 article · Prompt Engineering, Bias mitigation, Machine learning model evaluation, fairness testing, sentence completion

Prompted by GeekWire to create three separate children’s stories about animals — a bear, a bird and a cat — Google Gemini Storybook rendered three male characters. A new UW study dives much deeper into AI models’ gendering of animal characters. (Google Gemini Images) A bear is…

Microsoft extends zero trust deeper into enterprise AI

2 outlets · 2 articles · 3 total since August 5, 2026 · Zero Trust, microsoft, DevSecOps, security posture assessment, Zero Trust for AI Agents

Microsoft expanded its Zero Trust for AI strategy with updates to the Zero Trust Assessment tool and the Zero Trust Workshop. The additions help organizations assess security posture, prioritize remediation, and apply zero trust principles to AI agents and AI-assisted software…

More coverage: news.microsoft.com