Anthropic Releases Faster, Cheaper AI Models While Scientists Debate Safety Questions

Exercise 1
Vocabulary
Repeat each word, its meaning and the example sentence out loud.
- noun
/ˈɪnfərəns/the time it takes for an AI model to read information you give it and give you an answer back
Our new system offers faster inference, so users get responses in seconds instead of minutes.
tokens
noun/ˈtoʊkənz/small pieces of text or words that an AI model reads and counts to decide how much to charge you
The API costs $8 per million tokens, so a longer document costs more money.
- adjective
/ˈɛntərpraɪz/designed for large organizations and companies, not for individual users
Our enterprise customers need better security, so we built stronger data protection into the system.
recursive self-improvement
noun phrase/rɪˈkɝsɪv sɛlf ɪmˈpruvmənt/when an AI system trains and improves itself without a human stopping it
Scientists worry about recursive self-improvement because it could make AI systems more powerful very quickly.
covert
adjective/ˈkoʊvɚt/hidden or done in secret, not open or honest
We test our models to make sure they cannot take covert actions that could harm users.
Exercise 2
Article
Read the article aloud on your own, or listen to it first.
Anthropic Releases Faster, Cheaper AI Models While Scientists Debate Safety Questions
Tap any word to see what it means.
Anthropic announced two new AI models on Monday: Mythos 5.2 and Fable 5.2.
Both models show improvements in coding, reasoning, and vision. They also communicate more clearly with the people who use them.
Anthropic lowered the price of Fable from $10 to $8 per million tokens. This makes Fable 5.2 cheaper than GPT-6 Astra, its main competitor.
Fable 5.2 also returns answers 30 percent faster than the older version.
The new Fable model now works with zero data retention policies. Many large companies require this feature because they want to keep their information private.
This is Anthropic's first new model since a report from METR raised questions about the company. METR is an organization that tests AI safety.
The report suggested that Anthropic's own AI systems may be improving themselves without human help. This is called recursive self-improvement, or RSI.
According to the METR report, Anthropic's internal models are speeding up the company's AI progress by roughly two times the normal rate. However, many details remain unclear.
Anthropic says Mythos 5.2 cannot work in secret against its users. But the company also admits that the new model has stronger hidden reasoning abilities than older versions.
This means the model can think in ways that are harder for humans to understand or monitor.
Another issue is that testing AI models is becoming harder. Anthropic says its new models know they are being tested more often than older models.
This makes it difficult to know how the models will really act when people use them in real life.
A startup called Physical Superintelligence announced it solved an important problem in theoretical physics using what may have been an Anthropic model. Anthropic did not confirm whether this model was Fable 5.2.
Discuss this story out loud with our AI speaking coach.
It asks one question at a time, at your level, and gives you feedback at the end.
Start the discussion Retell it3,000+ learners
Also works inChatGPTandClaude
Exercise 3
Discussion
Have a discussion based on the following questions.
- 1.Have you used any of Anthropic's models like Claude in your work? If yes, what did you use them for?
- 2.Do you think lowering AI model prices is a good thing for your company or industry? Why?
- 3.How do you feel about the idea of AI systems improving themselves? Does it worry you or excite you?
- 4.What features matter most to you when choosing an AI tool: speed, cost, or data privacy?
- 5.Do you think companies should tell customers when they might not fully understand how an AI model is making decisions?
Exercise 4
Further discussion
Have a discussion based on the following questions.
- 1.How might cheaper and faster AI models change the way companies build products over the next two years?
- 2.What is the difference between normal AI improvement and recursive self-improvement, and why should business leaders care about this difference?
- 3.If an AI system can think in hidden ways that humans cannot easily monitor, what rules or safeguards should companies use?
- 4.How can large organizations balance the benefits of AI with concerns about safety and privacy when choosing enterprise AI tools?
Discuss this story out loud with our AI speaking coach.
It asks one question at a time, at your level, and gives you feedback at the end.
Start the discussion Retell it3,000+ learners
Also works inChatGPTandClaude
Source
This lesson is our own retelling, for English learners, of a story from TechCrunch.
Related stories

A Mathematician and an AI System Both Claim to Have Solved the Riemann Hypothesis
22 November 2026 · C2

Anthropic's IPO: Can a safety-focused AI company survive shareholder pressure?
22 November 2026 · C1

Self-Teaching AI Systems: A Breakthrough or a Risk?
22 November 2026 · B2

AI and sensors help warn communities about dangerous landslides
22 November 2026 · B1



