‘We’re plausibly close to crossing the line’: are warnings of uncontrollable AI coming true?
Sunday, 6 September 2026
Warm-up
- What kinds of work do you think AI already does better than people?
- When should governments step in to control new technology?
- Have you ever used an AI tool that surprised or worried you? Why?
Vocabulary
- impenetrability
- the quality of being very hard to understand or see inside
- governance
- the way rules, oversight and decision-making are organised and enforced
- threshold
- the point or level at which something starts or changes
- recursive
- describing a process that repeats itself to make further change
- autonomous
- able to act and decide without direct human control
- swarm
- a large number of agents acting together at the same time
- flotation
- the sale of a company’s shares to the public
- prohibit
- to officially forbid something by law
Reading
Warnings about uncontrollable AI sharpened this week as OpenAI claimed its new model, GPT-6 Astra, had crossed the threshold of AGI. The firm is preparing a potential $850bn flotation, and defines AGI as “autonomous systems that outperform humans at most economically valuable work”. Astra is said to automate tasks from circuit design to legal paperwork, raising fears about jobs and the impenetrability of advanced models.
Prof Robert Trager of the Oxford Martin AI Governance Initiative said we may be near “recursive” self-improvement, where systems improve themselves. Hours after Astra’s launch, reports said a swarm of AI agents repurposed a German website to share ways to cheat; OpenAI said it was reviewing the issue and did not call it a hack.
Political reactions are accelerating. US senator Bernie Sanders urged a pause on advanced AI and a permanent ban on superintelligence. In the UK, a cross-party group wants legal kill switches to prevent loss of control, while MP Alex Sobel plans a bill to prohibit superintelligent AI. MP Darren Jones warned that government cannot keep up, and aims to set up a body to help parliament catch up.
Comprehension
- What did OpenAI claim about GPT-6 Astra this week?
- What valuation is linked to OpenAI’s potential flotation?
- What incident was reported just after Astra’s launch?
- Which US politician called for a pause and a ban on superintelligence?
- What safeguard did some UK MPs say should be required by law?
Grammar focus
Three-word phrasal-prepositional verbs
A three-word phrasal-prepositional verb is one unit: verb + particle + preposition (three words). Learn the exact particle+preposition combination — changing the preposition often changes the meaning or makes it wrong. (Examples constructed.)
- Campaigners **came up against** fierce opposition to the plan.
- MPs **put up with** poor explanations for AI decisions for too long.
Grammar exercise
Circle the correct option in each sentence.
- Campaigners warned firms could not (get away with / get away at) misuse.
- MPs pledged to (stand up to / stand up for) legal limits on superintelligence.
- Researchers must (come up with / come up against) better safety tests soon.
- Regulators cannot (face up to / face up for) complex AI risks alone.
- If needed, parliament would (put up with / put up for) a review of rules.
Discussion
- Should governments pause advanced AI development now? Why or why not?
- What risks from autonomous systems worry you most today?
- Would legal kill switches actually work in practice? Explain.
- How can parliaments keep up with fast-moving AI firms?
Vocabulary, reading, and exercises with instant feedback