Last Week’s AI News #32

There was a moment when using AI was a choice.

That moment is fading. Because now it’s being built directly into the tools, systems, and rules businesses already operate in.

You don’t adopt it anymore. You operate inside it.

Here’s what you need to know from last week in AI:

• China blocks Meta’s $2B Manus deal
• Elon Musk vs OpenAI heads to court
• Google signs Pentagon AI deal amid internal backlash
• “Talkie” model trained only on pre-1931 data
• White House limits access to Anthropic’s Mythos
• Google brings Gemini into cars
• Pentagon adds AI firms but excludes Anthropic
• AI detects cancer years earlier in breakthrough study
• Everything else that happened in AI last week

CHINA BLOCKS META’S $2B MANUS DEAL

China vetoed Meta’s planned $2B acquisition of Manus, forcing both sides to unwind the deal despite already being operationally integrated.

Chinese regulators opened a probe into export controls and foreign investment rules shortly after the deal was announced. Authorities have now formally blocked the acquisition, treating the startup as strategically sensitive.

Meta claims the integration was already underway, with teams merged and operations aligned in Singapore.

Why does it matter for businesses?

AI talent is now being treated like chips, a controlled national resource.

This introduces a new layer of risk: even if deals are legally structured, geopolitical intervention can override them. Cross-border AI expansion is no longer just a business decision, it’s a political one.

ELON MUSK VS OPENAI HEADS TO COURT

Elon Musk took the stand in a $130B lawsuit against OpenAI, accusing the company of abandoning its original nonprofit mission.

Musk is seeking damages, leadership removal, and a reversal of OpenAI’s for-profit transition. OpenAI’s legal team argues the case is driven by competition, not principle.

The trial is expected to reveal internal communications and strategic decisions from the early days of OpenAI.

Why does it matter for businesses?

This case could redefine governance in AI companies.

It raises a fundamental question: can organizations pivot from mission-driven structures to commercial ones without consequences? The outcome may influence how future AI ventures are structured from day one.

GOOGLE SIGNS PENTAGON AI DEAL AMID INTERNAL BACKLASH

Google signed a classified agreement allowing its AI models to be used for government purposes, while hundreds of employees protested internally.

The deal reportedly gives the Pentagon broad usage rights, with limited restrictions. At the same time, Google has removed earlier commitments against military AI applications.

This follows similar moves by OpenAI and xAI.

Why does it matter for businesses?

AI labs are increasingly aligning with government priorities.

This creates tension between talent, ethics, and revenue. Companies building AI systems will need to navigate not just markets, but political alignment and internal trust.

“TALKIE” MODEL TRAINED ONLY ON PRE-1931 DATA

Researchers introduced Talkie, a model trained exclusively on data from before 1931 to study how AI behaves without modern internet influence.

Despite lacking exposure to modern tools, the model demonstrated reasoning capabilities, including writing functional code by generalizing from patterns.

The project avoids benchmark contamination and offers a new way to evaluate model intelligence.

Why does it matter for businesses?

Most AI systems today are trained on the same data, leading to similar outputs.

Alternative training approaches could unlock differentiation, and reveal that reasoning ability may not depend as heavily on modern data as assumed.

WHITE HOUSE LIMITS ACCESS TO ANTHROPIC’S MYTHOS

The U.S. government is pushing back on expanding private sector access to Anthropic’s Mythos model, citing compute constraints.

At the same time, internal divisions exist, with some officials advocating broader adoption and others resisting.

The situation unfolds alongside broader tensions between Anthropic and U.S. defense agencies.

Why does it matter for businesses?

Access to top-tier AI models is becoming restricted.

This creates a new competitive dynamic where availability, not just capability, determines advantage. Compute and policy are now strategic bottlenecks.

GOOGLE BRINGS GEMINI INTO CARS

Google is rolling out Gemini-powered AI into vehicles, replacing Assistant with a more advanced conversational system.

The system handles navigation, messaging, entertainment, and vehicle controls, while integrating with Google services.

Initial rollout begins in the U.S. across compatible vehicles.

Why does it matter for businesses?

AI is moving into physical environments.

While still early, this signals a shift toward embedded intelligence across devices, extending AI beyond screens into real-world interactions.

PENTAGON ADDS AI FIRMS BUT EXCLUDES ANTHROPIC

The Pentagon added multiple AI companies to its classified systems, including OpenAI, Google, and Microsoft, but excluded Anthropic.

This comes despite similar policy constraints across vendors and ongoing debates about Anthropic’s role.

The move highlights inconsistencies in how AI providers are evaluated.

Why does it matter for businesses?

AI partnerships are becoming selective and strategic.

Being excluded or included in key ecosystems will directly impact growth opportunities. Access to government contracts is becoming a major differentiator.

AI DETECTS CANCER YEARS EARLIER IN BREAKTHROUGH STUDY

Mayo Clinic introduced REDMOD, an AI system capable of detecting pancreatic cancer years before traditional diagnosis.

The model identified early-stage indicators in scans previously labeled as normal, significantly outperforming specialists.

It works by detecting subtle patterns invisible to the human eye.

Why does it matter for businesses?

AI is moving into high-stakes decision-making.

This isn’t just efficiency, it’s capability expansion. Industries like healthcare will be fundamentally reshaped as AI begins outperforming human experts in critical tasks.

EVERYTHING ELSE THAT HAPPENED IN AI LAST WEEK

• OpenAI expands via Amazon Bedrock: OpenAI made GPT-5.5, Codex, and Managed Agents available through Bedrock after restructuring its Microsoft deal. This lowers integration barriers and gives SMBs easier access to enterprise-grade AI via existing AWS stacks.
• Google adds file creation to Gemini: Gemini can now generate Docs, Sheets, Slides, Word, Excel, and Markdown files directly. This turns AI into a production tool, reducing manual work across everyday business tasks.
• Adobe launches Firefly AI Assistant (beta): Adobe opened its assistant for prompting multi-app Creative Cloud workflows with editable outputs. This simplifies creative production and reduces reliance on specialized roles.
• Meta opens ads platform to AI tools: Meta now allows third-party AI agents to manage ad campaigns via MCP server integrations. This signals a shift toward automated marketing operations accessible even to small teams.
• Anthropic expands creative workflow integrations: New connectors support tools like Blender, Adobe CC, and Fusion. This shows AI becoming a coordination layer across professional software, not just a standalone tool.
• Mistral launches remote coding agents: Mistral introduced cloud-based agents that run coding tasks in parallel sessions. This enables SMBs to execute development work without scaling engineering teams.
• Anthropic releases Claude Security (beta): Claude can now scan codebases for vulnerabilities and suggest fixes. This lowers the cost of security practices that were previously limited to larger companies.
• Cursor launches automated security review agents: Cursor introduced agents that continuously scan codebases and report results to Slack. This turns security into an automated, ongoing process instead of a manual one.
• Maryland bans AI-driven personalized pricing: A new law prohibits using AI to adjust grocery prices based on individual shopper data. This signals increasing regulation around algorithmic pricing strategies.
• Chinese court rules AI replacement doesn’t justify firing: A court ordered compensation after a company replaced a worker with AI. This suggests legal systems may slow down aggressive automation strategies.

AI is no longer just improving workflows.

It’s starting to define them. The companies that adapt will move faster with less effort.The ones that wait will find themselves working around systems they don’t control.

And that gap will only widen.

👉 Stay ahead,  get next week’s AI before everyone else.