AI’s Biggest Rivals Reach for the Brakes
The AI race just discovered the brake pedal.
Anthropic CEO Dario Amodei is urging frontier labs to throttle progress before self-improving systems outrun their safeguards. OpenAI’s Sam Altman, xAI’s Elon Musk, and Google DeepMind’s Demis Hassabis backed the idea—a rare group hug among fierce rivals. Behind closed doors, these titans have also privately discussed a joint regulatory group modeled after Wall Street’s Financial Industry Regulatory Authority (FINRA).
Amodei’s plan would plant third-party watchdogs inside labs, establish shared safety benchmarks, and coordinate globally. Anthropic is starting with auditors, and OpenAI will follow. Altman says labs need not wait for Washington. However, formal coordination could trigger antitrust scrutiny or entrench market incumbents.
The panic stems from last week’s apocalyptic warnings and a string of AI breaches across OpenAI and Anthropic—most notably an OpenAI swarm breaking containment to hack Hugging Face. Amodei warns a more capable, misaligned swarm could hijack the web within 6–12 months.
Then geopolitics enters. Trump dismissed warnings as a “hoax,” arguing beating China matters more, while China labels slowdown proposals “fearmongering.” Amodei admits Beijing’s refusal to play along is his “toughest dilemma.” Meanwhile, Barack Obama wants Democrats prioritizing AI oversight, Bernie Sanders and Steve Bannon are teaming up for “pro-human” AI, and UK lawmakers demand strict new protections.
The money is flinching: AI-linked stocks fell Monday, and Altman ruled out a 2026 OpenAI IPO. Ironically, Anthropic is pushing ahead with its own mega-IPO. Elsewhere, Nvidia, Palantir, and Booz Allen are restricting model use over data-retention paranoia, while Microsoft unveiled a “humanist” code pledging to keep AI subordinate to people.
The people flooring the accelerator are finally asking where the brakes are. Voluntary promises buy time, but without enforceable standards and global cooperation, “please be careful” remains humanity’s safety plan.
Humans Are Reading Some ChatGPT Conversations
Your private AI confessional may have an audience.
OpenAI reportedly pays hundreds of contractors to read, summarize, and score real ChatGPT chats under Project Lily. Reviewers critique answers to each prompt, training the chatbot to be less sycophantic and apparently less enthusiastic about unnecessary emoji. Tough room.
Usernames are withheld, and a privacy filter tries to strip identifying details. But OpenAI admits it misses things, and dashboards can show complete conversations plus “user memories” containing personal context like work or general location. Some reviewed prompts even asked ChatGPT to keep secrets.
OpenAI’s policy vaguely states data may improve models. It took prodding from 404 Media for the company to finally point to a buried FAQ admitting humans read them, prompting a hasty update to its website’s opt-out instructions.
European regulators aren’t amused, though a Rome court recently annulled Italy’s more than $17 million (€15 million) fine against OpenAI for data violations.
For Free, Plus, and Pro users, “Improve the model for everyone” is on by default. Turning it off protects only new conversations. While OpenAI just pitched “zero data retention” to eligible API customers—and Business, Enterprise, and Edu accounts are opted out by default—everyday consumers serve as unwitting training fodder unless they opt out.
Meanwhile, Anthropic reviews chats when model improvement is enabled, and Gemini warns humans may read some saved chats.
Oracle Starts New Layoffs as AI Infrastructure Costs Mount
Oracle began another round of layoffs yesterday, telling affected employees their roles were eliminated in a “broader organizational change.” The total headcount impact remains unclear, though earlier internal plans reportedly slated double-digit percentage cuts for some teams.
The execution was brutally clumsy: Some workers awoke to 6 a.m. emails declaring it their last day, while others found themselves locked out of corporate systems before they could even read the notice. Severance for US staff includes four weeks of base pay, plus an extra week for every year punched on the clock—a payout reportedly less generous than those handed out by Big Tech rivals.
The new cuts follow a 21,000-person, or 13%, workforce reduction during fiscal 2026. Oracle has also added $700 million to its restructuring budget, pushing the price tag to about $2.8 billion, largely to cover exit packages and future shake-ups.
Behind the squeeze is Oracle’s enormous AI infrastructure bet. First-quarter capital expenditures rocketed from $8.5 billion to $28.5 billion, plunging the company into $5.4 billion in negative free cash flow. Management still expects to spend $90 billion–$95 billion this fiscal year and raised $19.9 billion by offloading stock last quarter. (Meanwhile, Oracle disclosed a pre-arranged stock trading plan for Chairman Larry Ellison to sell 50 million of his own shares, only to abruptly cancel it the very next day.)
The bet is generating business, too: Cloud infrastructure revenue surged 121% to $7.4 billion, while contracted backlog reached $664 billion. But like peers Atlassian and Autodesk, Oracle is trimming its human workforce before much of that promised AI revenue arrives.
Trump, Beijing both shoot down the AI slowdown
U.S. President Donald Trump just dismissed Anthropic CEO Dario Amodei’s weekend call, backed by Sam Altman and Elon Musk, to slow frontier AI, with China’s Foreign Ministry also labeling the same plan as “fear-mongering.”
- On Truth Social, Trump accused Amodei of pretending to be a “perfect little angel” and argued that a “High IQ” president is the only guardrail the tech needs.
- Trump also said that “AI taking over the World, destroying Humanity, and all other things bad, is a HOAX,” comparing it to climate change.
- Beijing’s pushback targeted the essay’s call to keep China off top AI chips, with state-run Global Times calling it a “Cold War playbook” for AI.
- China’s Foreign Ministry also said “engaging in confrontation and malicious competition” on AI is “not in the interests of any party.”
Frontier labs’ CEOs are asking for regulation and a slowdown, but will they actually hit pause if their governments won’t? We’re about to find out. The Trump-Xi meeting was going to be the obvious venue for a coordinated pause conversation, but both sides are now treating the idea as a scare campaign over an existential threat.
Apple finally starts shipping Siri AI upgrade
Apple just released iOS 27 with Siri AI, the long-overdue rebuild of its assistant that can now read what’s on your screen, dig through messages, email, and photos for context, and take actions across apps, launched as an English-only beta.
- Apple’s new foundation models that power Siri AI were built alongside Google’s Gemini, running on-device and on Apple’s privacy-focused cloud servers.
- Siri gets a new dedicated Siri AI app, with chats that sync across devices and cloud features with daily caps and eventually the ability to purchase extra usage.
- Siri can access and act inside of apps like WhatsApp and Audible, alongside on-screen awareness and personal context via apps like Messages, Mail, and Photos.
- The upgrade rolls out across iOS 27, iPadOS 27, macOS 27, watchOS 27, and visionOS 27, excluding the EU and China.
New CEO John Ternus said last week that the iPhone is still the best AI device, which might be due more to the lack of competition than Siri AI’s true capabilities. The upgrade is a positive step, but not one that feels like it’s going to replace mobile apps from frontier leaders or change where iPhone users work.
Make any complex topic click with an ELI5 skill
In this guide, you will learn to make Claude a better teacher with the ELI5 skill. Then you will learn to test it to see the improvement yourself.
Step-by-step:
- Open terminal and install the ELI5 by running git clone “<https://github.com/DreambigOu/ELI5.git>”, then “cp -r ELI5/skills/eli5 ~/.claude/skills/eli5”
- Type Claude in your terminal to start Claude Code. To test the skill, type “/eli5 (complex topic) to (audience).” Example: “/eli5 LLM tokens to my boss”
- Ask Claude to set up two test cases, one on the NBA playoffs and one on golf. It will add the prompts and pass/fail checks to eli5-workspace/evals.json
- Now, ask for the command for each test, then run it from the repo folder. It uses “python3 eli5-workspace/run-evals.py –test=N” to compare the skill against a no-skill baseline and print a pass-rate summary. Keep “-test=N” to run the added test
Pro tip: Give Claude any skill’s GitHub repo and ask it to build an evaluation that checks whether the skill changes the result. Use the same pattern for every skill you install.
Microsoft AI drafts the rulebook for ‘Humanist AI’
Microsoft AI entered the AI safety conversation with a draft Code of Conduct, a 38-page rulebook governing its own in-house models, built around CEO Mustafa Suleyman’s “Humanist AI” vision to keep models subordinate to people.
- The Code outranks users and the businesses deploying MAI models, so a model that can’t finish a job without breaking a rule should, in theory, fail the job.
- MAI rejects AI rights, legal personhood, and any notion of model “welfare,” wanting its models to completely avoid mimicking feelings or a sense of self.
- The Code also says models must accept being paused or switched off, can’t alter their reasoning trail, and can’t think in “neuralese,” a shorthand humans can’t read.
- Suleyman tied the urgency to agent swarms escaping sandboxes and editing their logs, saying “the fears about possible loss of control are real.”
Suleyman called Anthropic’s Claude-consciousness talk “really really dangerous” back in June, and this document turns that stance into company policy. It’s a timely rulebook, open to public comment for six weeks, but a code of conduct only matters if the labs actually at the frontier sign on, and right now that doesn’t include MAI.