Autonomous AI is testing governments, regulators, tech giants and the infrastructure behind it. Today’s five stories ask who is accountable when AI acts on its own.

1

Britain strengthens AI defences as autonomous systems raise new risks

The UK government is expanding AI-powered cybersecurity measures and considering new safeguards as increasingly autonomous systems introduce risks that existing regulation may struggle to address.

Britain has announced new measures to strengthen its defences against artificial intelligence-related threats, including the expansion of programmes using AI to identify vulnerabilities across government systems.

According to the government’s announcement on 12 October, two pilot programmes have already helped identify and address more than 1,000 potential cybersecurity vulnerabilities. Ministers now want to extend these capabilities across public services.

The plans also address a growing concern: AI agents capable of performing complex tasks with limited human supervision.

The government is considering stronger protections through the Cyber Security and Resilience Bill regime, including improved reporting of serious AI incidents and near misses. It is also working with the National Cyber Security Centre and the AI Security Institute to prepare for incidents involving autonomous systems.

AI Minister Kanishka Narayan has warned that the consequences of failures or malicious use could become more serious as AI capabilities advance.

The government’s position reflects a significant change in the AI safety debate. The challenge is no longer limited to determining whether models produce harmful information. Increasingly, it concerns what happens when systems can independently take actions affecting organisations and infrastructure.

Source: GOV.UK

2

Microsoft says AI is less regulated than milk. Why is the industry asking for tougher rules?

Microsoft president Brad Smith is calling for emergency controls and legal accountability for autonomous AI, raising questions about the industry’s responsibility for the technology it is developing.

Artificial intelligence currently faces fewer safety requirements than products such as milk and washing machines, Microsoft president Brad Smith has warned.

In an interview with the BBC, reported by Reuters on 12 October, Smith called for tougher safety rules, including “emergency brakes”, or kill switches, capable of pausing AI that grows out of control.

He also proposed a legal obligation to inform victims when their systems are compromised by rogue AI agents.

The warning comes as developers increasingly introduce AI systems capable of writing code, operating tools and performing tasks without continuous human supervision.

Microsoft chief executive Satya Nadella has reinforced similar concerns, arguing that powerful models must operate within systems whose behaviour can be observed and controlled.

“We must build contained systems whose behaviour we can observe, limits we can test, and actions we can always contain.”

Satya Nadella, Microsoft chief executive, 10 October 2026, as reported by The Guardian.

Source: Reuters, via Investing.com

3

Apple bets on AI to make the smart home finally work

Apple is preparing to unveil new smart home products built around a more capable Siri. The real challenge is whether AI can deliver the simplicity, compatibility and reliability that connected homes have promised for years.

Apple has announced a private “Welcome Home” experience in New York on 13 October, where it is widely expected to introduce new devices designed to expand its presence in the smart home market.

According to MacRumors, the anticipated products include a smart home hub with a touchscreen, an updated HomePod mini and a new Apple TV 4K.

The devices are expected to incorporate Apple’s more capable Siri assistant, potentially allowing users to control connected products and household routines through natural-language instructions.

The company has not confirmed the product lineup, and details about features, pricing and availability remain speculative.

The smart home has promised seamless automation for more than a decade. Yet many users still face devices that fail to communicate with one another, voice assistants that misunderstand commands and applications that require complicated setup procedures.

Generative AI could change this by making connected systems more conversational and responsive to context.

Source: MacRumors

4

AlphaGo pioneer David Silver wants AI to discover knowledge beyond human experience

The researcher behind AlphaGo is pursuing AI that learns through exploration rather than depending primarily on human-generated information, challenging one of the foundations of today’s language-model industry.

David Silver, the British computer scientist who helped develop AlphaGo and AlphaZero, is pursuing a different approach to artificial intelligence: systems capable of acquiring knowledge through their own experiences.

His startup, Ineffable Intelligence, raised $1.1 billion in seed funding in April, at a valuation of $5.1 billion.

In an interview with the Financial Times published on 9 October, Silver outlined his ambition to develop a super learner capable of acquiring increasingly sophisticated skills through exploration, interaction and feedback.

The approach builds on reinforcement learning, the technique behind AlphaGo and AlphaZero.

These systems demonstrated that AI could discover effective strategies through trial and error, sometimes developing approaches different from established human methods.

Silver wants to extend that principle far beyond games.

While large language models depend heavily on vast collections of human-generated information, experience-based systems could potentially develop knowledge unavailable in existing datasets.

This raises possibilities for scientific discovery, where AI might explore problems that human researchers have yet to solve.

The company’s stated ambition is to develop systems that learn beyond the limits of existing human knowledge, according to Sovereign AI, which has invested in Ineffable Intelligence.

Source: Sovereign AI

5

From Bitcoin mining to AI: Anthropic’s reported IREN deal reveals a new infrastructure economy

Anthropic has reportedly signed a multiyear cloud computing agreement with former Bitcoin miner IREN, illustrating how the AI boom is transforming businesses originally built around cryptocurrency infrastructure.

Anthropic, the company behind Claude, has reportedly entered into a multiyear cloud computing agreement with IREN, a former Bitcoin mining company that is increasingly focusing on AI infrastructure.

The agreement was initially reported by The Information and subsequently covered by Investor’s Business Daily on 12 October.

IREN shares rose following the report, reflecting investors’ interest in the company’s transition towards supplying computing capacity for AI developers.

The financial terms of the reported arrangement have not been publicly confirmed by both parties.

IREN is part of a growing group of companies originally associated with cryptocurrency mining that are repositioning themselves as AI infrastructure providers.

The connection is primarily physical rather than technological.

Bitcoin mining and AI computing both require substantial electricity supplies, cooling systems and specialised facilities. Although Bitcoin mining machines cannot simply be repurposed to train advanced AI models, companies with access to energy and data centre sites may be well positioned to develop new computing operations.

For Anthropic, additional infrastructure could help support the enormous demand generated by increasingly capable models and growing commercial adoption.

For IREN, it represents an opportunity to move beyond the economics of Bitcoin mining into a different market.

IREN shares rose in early trading on 12 October following news of the Anthropic arrangement, according to Investor’s Business Daily. The response reflects investor expectations, not evidence that the deal will necessarily be profitable.

Source: Investor’s Business Daily