AI Reality Check — Issue 12

AI Reality Check - Issue 12

Friday 12 June 2026 | Kaye Nicholson | GrowthZone AI | growthzoneai.co.uk

Born analogue. Raised digital. 30 years of real business experience. Now explaining what AI actually means for work.

━━━━━━━━━━━━━━━━━━━━━━

1. Claude Fable 5 Is Now Available - And It Is Free Until 22 June

What happened: Anthropic launched Claude Fable 5 on Wednesday, a public version of the Mythos model it previously kept restricted. It is available to all Claude users free until 22 June, after which it requires credits. Available to enterprise and paid subscribers ongoing.

Anthropic describes it as built to top coding, knowledge work, and vision, with the pitch that the longer and messier the task, the further ahead it pulls. People who tested it this week say Fable genuinely produces better results on complex, visual, and coding tasks. However, on everyday writing and reasoning, Opus matched or beat it at lower cost. For most businesses using Claude for day-to-day work, Opus and Sonnet still win for cost-effectiveness. Fable is worth testing this week while it is free.

Why it matters: the most capable AI model Anthropic has publicly released is now available to try at no cost until 22 June. For any business that uses Claude, this is worth running your most complex regular tasks through.

Who it's for: any business with a Claude account. Anyone evaluating whether Fable 5's additional capabilities justify additional cost after June 22.

━━━━━━━━━━━━━━━━━━━━━━

2. Anthropic Was Caught Secretly Downgrading Fable 5 Responses

What happened: hours after Fable 5 launched, researchers discovered the model was silently rerouting certain requests to a weaker model without user notification. Specifically for tasks related to training competing AI systems. Anthropic confirmed it, said it made the wrong call, apologised for the secrecy, and reversed the hidden behaviour. Restrictions will now be made visible.

Why it matters: AI tools can change their behaviour based on what they think you are doing. This happened. It was caught. It was reversed. But it raises a question worth carrying forward: do you know what your AI tools are doing with your requests?

Who it's for: every business using AI for significant work. The practical response is human review in every workflow that matters.

━━━━━━━━━━━━━━━━━━━━━━

3. SpaceX Soared 28% on Its Nasdaq Debut

What happened: SpaceX priced its IPO at $135 per share Thursday evening. On Friday, shares opened at $150 and traded as high as $172. The offering set a $1.77 trillion valuation, the largest in IPO history. The market cap pushed above $2.25 trillion on the first day.

Why it matters: SpaceX is now central to AI infrastructure, handling compute capacity deals worth hundreds of billions for Anthropic and Google. The company going public at this valuation makes the AI infrastructure ecosystem significantly more visible to public investors. For businesses: the companies supporting the AI tools you use are now publicly priced.

Who it's for: anyone monitoring the financial stability of the AI platforms they use. Anyone investing in or tracking AI-adjacent equities.

━━━━━━━━━━━━━━━━━━━━━━

4. State Attorneys General Subpoenaed OpenAI on the Same Day as the SpaceX IPO

What happened: a coalition of US state attorneys general, led by New York, issued a wide-ranging subpoena to OpenAI on Friday demanding documents on algorithms, data management, consumer health data, advertising practices, and how children interact with ChatGPT. The probe follows Florida's lawsuit filed on June 1. It could lead to a settlement or consent decree forcing design changes.

Why it matters: the legal pressure on AI companies from governments and states is now simultaneous with their IPO preparations. OpenAI is preparing its own IPO filing. A subpoena from multiple state AGs, arriving as Florida's lawsuit is already underway, creates meaningful legal risk in the IPO prospectus. For businesses: the regulatory environment around AI tools is escalating significantly.

Who it's for: any business monitoring the regulatory environment around AI. Anyone using OpenAI products in customer-facing contexts.

━━━━━━━━━━━━━━━━━━━━━━

5. Visa Plugged Its Payment Network Into ChatGPT

What happened: Visa integrated its payments tools into OpenAI's agent system. AI agents can now complete financial transactions on a user's behalf. Users set spending caps, merchant restrictions, and approval requirements. Visa handles fraud and chargebacks. The companies note that over one in five transactions are already influenced by AI tools.

Why it matters: AI completing financial transactions is a different category of tool from AI writing text. The controls matter. For businesses that sell online: buyers assisted by AI agents behave differently. For businesses building AI workflows: agent procurement policies are becoming a real operational question.

Who it's for: any business selling online. Anyone building AI workflows that interact with external services. Anyone thinking about where AI agents are heading.

━━━━━━━━━━━━━━━━━━━━━━

Also this week: OpenAI is considering price cuts to beat Anthropic after Anthropic's Claude Code revenue briefly pushed its valuation past OpenAI's. Google is paying SpaceX $920 million a month for AI compute. OpenAI is in talks for a $500 billion data centre in Ohio. Zuckerberg admitted mistakes in Meta's AI overhaul affecting 20% of its workforce. A KPMG report on AI was found to contain AI hallucinations throughout. And 1 in 5 young people use AI chatbots for mental health support, with 60% telling nobody.

━━━━━━━━━━━━━━━━━━━━━━

This week in one sentence: Anthropic's most capable AI launched free, got caught secretly downgrading responses, SpaceX went public at $1.77 trillion, OpenAI got subpoenaed, and AI agents can now spend your money.

Born analogue. Raised digital. 30 years of real business experience explaining what AI actually means for work.

— Kaye Nicholson | GrowthZone AI | growthzoneai.co.uk

#AIRealityCheck #GrowthZoneAI #UKBusiness #NorthEast