7 June 2026: Lockdown Mode makes AI security visible (AM)

This morning’s AI news is less about another benchmark race and more about the controls around everyday use. OpenAI is putting prompt injection risk in front of ordinary ChatGPT users, Apple is heading into WWDC with pressure to make Apple Intelligence feel useful, and US policy signals are still shifting around the largest AI companies.

OpenAI says Lockdown Mode is now available to all logged in ChatGPT users, turning prompt injection from a specialist security term into a visible product setting. In its ChatGPT release notes, OpenAI says the optional setting limits access to the web and external services to reduce data exfiltration risk from malicious instructions hidden in webpages or uploaded files. TechCrunch reported that the mode disables or restricts live browsing, deep research, agent mode, file downloads and some web derived image support.

The important caveat is that OpenAI is not presenting the setting as a complete shield. According to the company, prompt injection can still affect responses through cached web content or uploaded files. For small firms, the practical message is simple: connected AI is useful because it can reach out to tools and data, but that same reach creates risk. If your team uses ChatGPT for supplier research, client files or internal documents, Cristoniq’s guide to prompt injection is the background check to read before treating agent features as routine office software.

OpenAI is also beginning to roll out ads for Free and Go users in the UK, while paid business and education tiers remain ad free. The same release notes say Plus, Pro, Business, Enterprise and Education plans will not carry ads. That makes the UK one of the first markets where everyday users may see ChatGPT’s commercial model change directly inside the product, not just in pricing pages or investor briefings.

For readers, the question is not whether ads are surprising. Free services usually need a business model. The point is what advertising does to trust in an assistant that people increasingly use for work, research and personal admin. If ads appear beside answers, OpenAI will need to make the boundary between paid placement, product recommendation and neutral response unmistakable. Businesses comparing AI plans should treat ad free access, admin controls and data handling as part of the cost, not as small print after the model choice.

Apple’s WWDC starts on 8 June with AI expectations unusually high, because Apple Intelligence still has to prove it can become a daily assistant rather than a feature label. Apple’s UK newsroom says WWDC26 will include AI advancements, software updates and developer tools. TechCrunch’s preview says Siri and Apple Intelligence are the areas to watch, including whether Apple can make its assistant more useful without losing its privacy centred positioning.

This is a preview story, so it needs caution. Apple has not announced the actual WWDC AI features yet. Still, the timing matters for consumers and small firms because Apple controls the default computing environment for millions of iPhone, iPad and Mac users. If Apple opens more useful on device AI tools to developers, the best AI feature for many people may not be a separate chatbot subscription. It may be something built into the apps they already use. Cristoniq’s explainer on on device AI covers why that distinction matters for privacy and speed.

Sriram Krishnan is leaving his White House AI adviser role at the end of June, according to TechCrunch, keeping attention on how the Trump administration’s AI policy is being shaped outside formal agencies. The report says Krishnan plans to build institutions focused on issues including energy, data centres and AI benefits for Americans and allies. It also notes that he had been involved with an AI Action Plan that prioritised data centre construction over regulation and safety.

For UK readers, this is not a direct rule change. It is a signal about the direction of US AI policy, and US policy still affects the tools, cloud capacity and compliance posture used by British companies. If Washington keeps prioritising infrastructure and competition over tighter safety obligations, UK firms buying American AI services may see faster product rollout but more responsibility pushed onto customers. The useful thing to watch is whether policy momentum moves through government agencies, outside institutions or direct deals with the biggest labs.

Here is everything else worth knowing from today’s AI news.

• The Trump administration may explore taking equity stakes in major AI companies; TechCrunch reported President Trump had discussed deals where the public could benefit from AI company success, so treat this as an early policy signal rather than a settled transaction.
• OpenAI has set ChatGPT retirement dates for older paid models; according to its release notes, o3 leaves ChatGPT on 26 August 2026 and GPT-4.5 leaves ChatGPT on 27 June 2026, with no API change announced.
• ChatGPT active session controls are rolling out; OpenAI says users can review first party sessions and sign out of devices they do not recognise, which matters as AI accounts become more valuable work targets.
• Apple’s developer tools are part of the WWDC AI story; Apple’s own event note says developers will get updates on Apple Intelligence, machine learning and platform tools, not only consumer Siri features.

The thing to watch next is whether AI companies keep turning abstract risk into concrete settings. The more assistants browse, download, remember and act, the more ordinary users will need controls they can understand before something goes wrong.

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s daily guide to developments in artificial intelligence, published every morning and evening.