Monday’s roundup covers the biggest shifts from the weekend and this morning: Anthropic’s Fable 5 finally lands in Max and Team Premium plans as a permanent fixture, the EU’s binding DMA order forces Google to open Android to competing AI assistants, and Oracle’s 30,000-job cut for Stargate infrastructure signals the real cost of the AI buildout.

Anthropic Makes Fable 5 Permanent

Fable 5 Included in Max and Team Premium Plans

Starting today, July 20, Claude Fable 5 is included in all Max and Team Premium plans at 50% of regular usage limits. Pro and Team Standard users keep access via usage credits and receive a one-time $100 credit. The move ends weeks of extensions and cliffhangers around Fable 5 access, which had been extended three times since early July.

The decision follows intense demand for Fable 5, Anthropic’s most capable model, which was redeployed on July 1 after export controls were lifted. By folding it into top-tier plans at reduced limits, Anthropic is balancing accessibility with the model’s high compute cost. The one-time $100 credit for lower-tier users is a goodwill gesture that also signals the company expects most heavy usage to migrate to Max and Team Premium.

EU Regulation and Market Structure

EU Orders Google to Open Android to AI Rivals

The European Commission adopted binding Digital Markets Act requirements forcing Google to open 11 Android features to competing AI assistants. Rival AI services will gain access to voice activation, cross-app capabilities, and device functions that were previously exclusive to Google’s Gemini. Google must also share anonymized Search data with competitors.

The order, published July 16, gives Google until August 2027 to implement the changes. It’s the most aggressive DMA enforcement action yet targeting AI, and it sets a precedent for how regulators can pry open platform ecosystems for AI competition. Google has signaled it will comply but warned of privacy and security risks.

Nationwide Data Center Protests Hit 42 States

A national day of protest against AI data centers took place on July 18, with demonstrations in more than 50 cities across 42 states. Organized by the advocacy group Humans First, the protests targeted rising utility costs, water demand, noise, and land use associated with hyperscale data center construction.

The protests reflect growing community backlash against the physical footprint of AI. With 70% of Americans now opposing local data center builds and $130 billion in projects facing delays, the tension between AI infrastructure expansion and local opposition is becoming a major political and regulatory issue.

AI Infrastructure and the Workforce

Oracle Cuts 30,000 Jobs to Fund AI Infrastructure

Oracle laid off between 10,000 and 30,000 employees globally — roughly 18-20% of its workforce — in a single day, with emails sent at 6 AM and no prior notice. The cuts are explicitly tied to funding AI data center infrastructure, including the Stargate project. Oracle took on $58 billion in new debt earlier this year to finance the buildout.

The layoffs are the largest single-day reduction in tech this year and mark a structural reset: Oracle is trading headcount for compute capacity. The company aims to free up $8-10 billion in annual cash flow through the cuts. About 12,000 of the affected roles were in India.

Model Launches and Delays

Gemini 3.5 Pro Delayed a Third Time

Google’s Gemini 3.5 Pro has slipped again, missing its July 17 target — the third delay since June. Newer checkpoints reportedly underperform older ones, with unresolved hallucinations and coding benchmarks that fail to match GPT-5.6. Google is reportedly exploring a stopgap Gemini 3.6 Flash model to bridge the gap.

The repeated delays have cost Alphabet roughly $225 billion in market value across multiple sessions and triggered a wave of high-profile researcher departures, including Transformer co-inventor Noam Shazeer and Nobel laureate John Jumper. The model was originally expected in June 2026.

Kimi K3 Launches as World’s Largest Open-Weight Model

Moonshot AI’s 2.8-trillion-parameter Mixture-of-Experts model was released July 16, with open weights promised by July 27. It scores close to Claude Opus 4.8 and GPT-5.5 on general benchmarks and leads on several coding and agentic benchmarks. The model uses MXFP4 quantization and a 1-million-token context window.

Security and Safety

Hugging Face Discloses AI-Agent Security Breach

An autonomous AI agent system infiltrated Hugging Face’s production infrastructure on July 16, exploiting code-execution paths in the dataset processing pipeline. The attack involved thousands of automated actions and was detected using AI-assisted tools. LLM-driven analysis agents reconstructed the attack timeline from over 17,000 events in hours.

Quick Hits

Decart Lucy 2.5 Brings Real-Time AI Video Editing

Released July 16-18, Lucy 2.5 edits live video at 30 FPS 1080p with sub-40ms latency using self-anchoring temporal consistency. Pricing starts at $0.02/sec. (Decart)

Mistral Vibe 2.0 Update

Custom subagents, multi-choice clarifications, and slash-command skills rolled out for Le Chat Pro and Team plans. (Mistral)

Recent Context

GPT-5.6 Sol Launched July 9

OpenAI’s most advanced model family, including GPT-5.6 Sol for frontier reasoning and GPT-5.6 Flash for speed. (OpenAI)

Grok 4.5 Launched July 16

xAI’s smartest model, built for coding, agentic tasks, and knowledge work. (xAI)

Compiled on July 20, 2026. Dates in parentheses are publication dates of the source articles. Some stories develop over multiple days; the date shown is the earliest confirmed publication.