OpenAI has begun rolling out Lockdown Mode to ChatGPT, a new security setting designed to block attackers from stealing data through prompt injection attacks. The feature disables live web browsing, agent mode, deep research, image retrieval, Canvas networking, and file downloads. It is available to logged-in users across Free, Go, Plus, Pro, and self-serve ChatGPT Business plans.

Prompt injection remains what OpenAI calls a “frontier” problem affecting all large language models. The attack works by hiding malicious instructions in content the model processes, such as a webpage or uploaded file. If the model follows those instructions, it can be tricked into sending sensitive data to an attacker-controlled server.

Lockdown Mode does not stop injections from happening. A malicious payload embedded in a cached webpage or uploaded PDF can still influence the model’s behaviour. What it does is shut down the outbound pathways an attacker would use to exfiltrate the data. No live browsing means no network requests to external servers. No image retrieval means no pixel-based data channels.

“Lockdown Mode is designed to substantially reduce the risk of prompt injection-based data exfiltration, but it does not guarantee that data exfiltration cannot happen,” OpenAI said. “Risk may remain through enabled Apps, unforeseen combinations of capabilities, or newly discovered techniques.”

The trade-off is significant. With Lockdown Mode on, ChatGPT loses most of what makes its agent and research features useful. Live browsing drops to cached content only. Agent mode is gone entirely. Deep research is disabled. It is, as OpenAI acknowledges, “not intended for everyone.”

The feature arrives as prompt injection attacks on AI agents have become a growing concern. Security researchers have demonstrated hijacks against agents from Anthropic, Google, and Microsoft via their GitHub Actions integrations. All three paid bug bounties but published no public advisories. The underlying weakness is fundamental: LLMs cannot reliably separate data from instructions.

Lockdown Mode and Developer Mode cannot be used simultaneously. Turning one on disables the other. OpenAI also launched a separate session management feature that lets users review active ChatGPT sessions and log out of individual devices if they spot unauthorised activity.

The feature is a pragmatic concession. OpenAI is not claiming to have solved prompt injection. It is accepting that the problem persists and offering users a way to reduce their exposure by giving up functionality. For anyone handling sensitive data in ChatGPT, that trade-off is worth making. For everyone else, the expanding agent ecosystem and its growing attack surface mean the risk is only increasing.

OpenAI announced a new feature that it says will provide additional protection from prompt injection attacks, where malicious chatbot instructions are hidden in webpages and other content sources.

Among other things, Lockdown Mode will disable live web browsing (so you can only access cached content), the retrieval and display of images from the web (you can still generate images), deep research, and agent mode.

The company says that even with Lockdown Mode turned on, ChatGPT could still be vulnerable to prompt injections — which could, for example, “appear in cached web content or in an uploaded file, and could still affect the behavior or accuracy of a response.”

But the goal is to reduce the likelihood that sensitive data gets shared in the process.

“Lockdown Mode is not intended for everyone,” OpenAI says. “It is designed for people and organizations that handle sensitive data and want stricter protection from data exfiltration risks related to prompt injection.”

The company says it’s currently rolling Lockdown Mode out to self-serve ChatGPT Business accounts, as well as eligible personal accounts.

Google may have just accidentally shown everyone where Gemini is headed next. According to TestingCatalog, a new Troubleshooting mode has quietly appeared inside the Gemini model picker menu for some users.

It sits alongside existing options like Gemini 3.5 Flash and 3.1 Pro, which are the standard AI models you already switch between in the app.

GOOGLE : A new Troubleshooting mode has been spotted on Gemini.

In this mode, Gemini will explain troubleshooting process via text responses and interactive widgets. Even though it is working and available, it still looks like an unintended release and might get reverted… https://t.co/FWQLelYXju pic.twitter.com/Y73PJb7y1e

— AI News | TestingCatalog (@testingcatalog) June 4, 2026

What does the Troubleshooting mode in Gemini actually do?

Rather than giving you a wall of text to read, the Troubleshooting mode guides you through a problem step by step using a mix of text responses and interactive widgets.

For example, if you tell Gemini your car will not start, it might identify common causes like a dead battery and then present you with symptom options to tap, such as “clicks or silent,” to help narrow down the issue faster. It is a more structured, guided experience than asking Gemini a question in regular chat mode.

How is this different from just asking Gemini normally?

That is a fair question, and the answer comes down to how the mode is tuned under the hood. Redditors who got early access suggest it runs on a lower temperature setting, which means it sticks closely to the problem at hand and skips the conversational filler.

Its responses are reportedly focused on diagnosis and practical fixes rather than general information. Google has not officially announced the feature, and it remains unclear whether this is a planned rollout or an internal test.

For now, the Troubleshoot feature appears to be an unintended release, meaning Google likely flipped it on by mistake, and could pull it back at any time. More details are expected in the coming weeks.

If you find Gemini’s new Troubleshooting mode exciting, there is a lot more happening with the assistant right now. Google just unveiled Gemini Spark, a 24/7 AI agent that handles your tasks in the background. On the flip side, free users may soon face stricter weekly usage caps.

After a very profitable decade on Microsoft’s board, Reid Hoffman is stepping down, the company announced Thursday. Hoffman joined the board after Microsoft bought his company LinkedIn for $26.2 billion in 2016.

Hoffman was on Microsoft’s board when it invested its first $1 billion into OpenAI in 2019. Hoffman was one of OpenAI’s original investors and served on the model maker’s board until he stepped down in 2023, citing too many potential conflicts of interest to continue. He was also on Microsoft’s board when the tech giant entered into one of those non-acquisition, acqui-hire deals for $650 million with his AI startup Inflection AI. Microsoft hired Inflection co-founder Mustafa Suleyman through that deal.

Hoffman said on a recent episode of his “Possible” podcast, while talking with Microsoft CEO Satya Nadella, that he’s ready to go “founder mode” with his latest AI startup, Manus. Manus is a drug discovery company that raised over $50 million through a couple of seed rounds last year. Hoffman is an investor, as is General Catalyst.

Hoffman is cited as a co-founder of Manus and chairman of the board, not the CEO, though. That job belongs to Dr. Siddhartha Mukherjee, a physician, biologist, and Pulitzer Prize-winning author of the 2011 book “The Emperor of All Maladies: A Biography of Cancer.”

Still, Hoffman said he’s excited to give Manus more attention.

“One of the things I realized over the last month was that, we’re seeing such progress with Manus. I need to get back to founder mode,” he said. He believes the startup is making progress on “Move 37” AI, meaning AI that supersedes human creativity in chemistry, especially to combat various cancers, he added.