AI Daily Brief: 5 June 2026
5 June 2026
Quick Read: Anthropic filed a confidential S-1 with the SEC at a $965 billion valuation, the highest ever for a startup. Co-founder Jack Clark warned BBC Newsnight that AI code is already 80% self-written and could reach full autonomy within two years. UK banks including Lloyds, HSBC and Nationwide are gaining access to OpenAI's GPT-5.5 Cyber after being blocked from Anthropic's Mythos. US lawmakers unveiled the Great American AI Act, which would block states from regulating AI model development for three years.
Two stories dominate this morning: Anthropic has confidentially filed for an IPO at a near-trillion-dollar valuation, while its own co-founder is urging governments to build a 'brake pedal' for AI development. Meanwhile, UK banks are scrambling to access frontier AI cyber tools, and US Congress has released a bill that would freeze state-level AI regulation for three years.
Anthropic files for IPO at near-trillion-dollar valuation
Anthropic confidentially submitted a draft Form S-1 to the US Securities and Exchange Commission on 1 June 2026, formally beginning the process of going public. The filing follows the company's $65 billion Series H funding round, which valued Anthropic at $965 billion - making it the most highly valued startup in the world, surpassing OpenAI's $852 billion valuation for the first time.
The company has engaged Goldman Sachs and Morgan Stanley as underwriters. Market analysts suggest a listing could arrive as early as autumn 2026, though the SEC review process makes a pre-July debut unlikely. Amazon and Alphabet, both significant investors, stand to benefit substantially from any public listing. The IPO would be among the most valuable stock market debuts in history.
Our take: Anthropic going public changes the game for the entire AI industry. A listing at near-$1 trillion forces every serious investor to take a position on AI as a category - not just as a tech subsector. For UK businesses, this signals that frontier AI is now a permanent feature of the corporate landscape, not a speculative experiment. Boards that have been sitting on the fence about AI strategy will find it harder to justify inaction when the makers of Claude are valued alongside the world's biggest companies.
Anthropic co-founder: AI needs a 'brake pedal' - and we don't have one
Jack Clark, co-founder of Anthropic, told BBC Newsnight this week that the AI industry is accelerating toward a point where systems could develop themselves without human input - and that no mechanism exists to slow it down. 'Right now, it's like the AI industry has a gas pedal, but it doesn't have a brake pedal,' Clark said. Already, 80% of the code in Anthropic's Claude was written by the AI system itself. Getting to 100% is possible within two years, Clark warned.
Clark drew a parallel to the early oil industry, calling for a regulatory framework that gives society confidence in AI systems regardless of the personalities running the companies building them. The warning came the same week Anthropic filed for its IPO - a tension Clark did not address directly. Notably, Anthropic also welcomed a relatively hands-off executive order on AI from President Trump, and has not indicated it will pause its own research.
Our take: Clark's warning matters more because of who is saying it, not just what is being said. This is not a campaigner or a regulator - this is the co-founder of one of the two most powerful AI labs in the world, saying publicly that his own industry lacks a functioning safety mechanism. UK business leaders should take note: if the people building these systems are calling for brakes, the window for governments to establish meaningful oversight frameworks is now, not after the next capability jump.
UK banks get OpenAI cyber AI access as Anthropic expands Mythos globally
Nine major UK banks - including Lloyds Banking Group, HSBC, Nationwide Building Society, NatWest Group, and Santander UK - are gaining or expanding access to OpenAI's GPT-5.5 Cyber model after some were previously blocked from accessing Anthropic's Claude Mythos Preview. Both models are designed to find hidden software vulnerabilities. The UK's AI Security Institute assessed that the two tools reached 'a similar level of performance' on the tasks it evaluated.
In a parallel development, Anthropic is significantly expanding Project Glasswing - its controlled access scheme for Mythos - from around 50 organisations to 150 more vetted partners across 15 countries. Speaking at Infosecurity Europe, the NCSC's director of operations Paul Chichester said China was approximately eight months behind on frontier AI cyber capabilities. Security expert Gunter Ollmann of Cobalt warned delegates to 'prepare for the son of Mythos' as Chinese and other competitors close the gap. Bank of England Governor Andrew Bailey has warned that cyber resilience 'can no longer be assumed' given the pace of AI-powered vulnerability discovery.
Our take: The race between AI-powered attack and defence tools is moving faster than most enterprise security teams appreciate. The fact that both Anthropic and OpenAI are now actively distributing frontier cyber AI to financial institutions - under regulatory pressure - tells you something important: regulators have already concluded that not having access to these tools is more dangerous than having it. For UK organisations outside the banking sector, the question is when, not if, access to AI-powered vulnerability scanning becomes a compliance expectation rather than a competitive advantage.
US Congress drafts bill to freeze state AI laws for three years
A bipartisan pair of US lawmakers - Republican Jay Obernolte and Democrat Lori Trahan - have released a discussion draft of the Great American Artificial Intelligence Act of 2026. The bill's most contentious provision would prevent US states from regulating the development of AI models for three years, effectively overriding laws already passed in California, New York, and Illinois. States would retain the right to regulate how AI is used within their borders, but not how it is built.
The bill would also require top AI developers to disclose safety and security risks associated with new models. Critics, including Public Citizen, argue the bill strips states of the authority to protect consumers, workers, and children while deferring to federal oversight frameworks that do not yet exist. AI legislation is simultaneously moving rapidly at state level: Illinois has sent five AI bills to the governor, California moved 30 AI bills through second-chamber committees this week, and Vermont has banned AI therapy bots.
Our take: This matters for UK businesses even though it is US legislation. A three-year federal preemption of state AI safety laws would be a significant gift to frontier AI developers, effectively locking in a lightly regulated development environment during the most critical years of capability growth. If it passes, the US will have chosen speed over safeguards at the federal level - putting pressure on UK and EU regulators to either match that pace or accept that the most capable AI systems are built under minimal constraint. Watch how the UK's AI Action Plan responds.
Quick Hits
- ChatGPT's updated 'Dreaming' memory system now builds coherent user profiles from conversations instead of saving disconnected bullet points, making long-term personalisation significantly more effective.
- Google DeepMind's Gemma 4 12B is being pushed toward local, laptop-ready multimodal intelligence - a significant step toward capable AI that runs entirely on consumer hardware without a cloud connection.
- Vermont has banned AI-powered therapy bots under new consumer protection legislation, becoming one of the first US states to explicitly restrict AI in mental health settings.
- Security experts at Infosecurity Europe warned enterprise teams to prepare for 'son of Mythos' - a wave of competing frontier AI vulnerability discovery tools from Chinese and other developers expected to reach comparable capability within months.
Frequently Asked Questions
How often is the AI Daily Brief published?
Every morning at 7:30am UK time, covering the previous 24 hours of AI news from over 30 sources.
How are stories selected?
UK-relevant stories are prioritised first, then by business impact and practical implications for UK organisations adopting AI.
Why should business leaders follow AI news?
AI is moving faster than any technology in history. Staying informed is essential for making smart decisions about AI investment, adoption, and governance.