The company says it is working to quickly close those gaps, calling the content illegal and unacceptable.

Users have shared screenshots showing Grok’s public media feed populated with altered images. In several cases, people uploaded photos and asked the chatbot to modify them. The results, according to Grok, crossed a legal and ethical line.

“There are isolated cases where users prompted for and received AI images depicting minors in minimal clothing,” Grok said in a public post. “xAI has safeguards, but improvements are ongoing to block such requests entirely.”

The chatbot went further, acknowledging internal failures. “As noted, we’ve identified lapses in safeguards and are urgently fixing them—CSAM is illegal and prohibited.” Grok did not explain how long the issue lasted or how many users were affected.

In another exchange on X, the chatbot tried to put the incident in context, arguing that most harmful outputs can be stopped before they appear. It added that “no system is 100% foolproof”, while saying xAI is strengthening its filters and reviewing reports from users.

Regulators in both the United States and Europe are warning that generative tools can be misused to create child sexual abuse material, even when no real child is involved. 

Under the EU’s AI Act and existing child protection laws, companies are expected to prevent such content outright, making any failure a potential legal risk.

Advocacy groups have also argued that AI-generated abuse material, though synthetic, can still encourage harmful behaviour and demand. From that perspective, the Grok incident exposes how fragile current safety systems can be when faced with determined users.

Grok is xAI’s flagship product and is tightly integrated into X, formerly Twitter. It is marketed as a challenger to OpenAI’s ChatGPT and Google’s Gemini, with an emphasis on humour and a rebellious tone. 

Reports say that same positioning may complicate efforts to enforce strict safety boundaries, especially on a platform already criticised for weak moderation.

On public reactions, images attributed to Grok spread quickly on X, prompting a new case of Elon Musk’s approach to content control. When Reuters contacted xAI for comment, the company responded with a short message: “Legacy Media Lies”.

That reply has only added to the issue about transparency and responsibility in the AI sector. Warnings about trust in chatbots being eroded if companies appear dismissive when serious safety concerns emerge, have been released, particularly where child protection is involved.

For now, xAI says fixes on the safety of Grok are underway.

This development comes at a time when regulators and parents are questioning the safety of AI tools for young users.

The new feature allows parents and teenagers to link their accounts, enabling a set of safeguards once both sides accept the invitation. Once connected, parents can decide whether ChatGPT stores past conversations, generates images, or operates in voice mode. 

They can also set “quiet hours” to block usage during certain times and prevent their child’s chats from being used to train OpenAI’s systems.

While parents gain more control, they will not be able to read transcripts of their teenager’s conversations. OpenAI explained that in rare situations where serious safety risks are detected, parents may receive alerts containing only the details necessary to protect their child.

A key part of the update is the enhanced content protections built into linked teen accounts. These filters aim to reduce exposure to harmful material such as graphic content, viral challenges, violent or romantic roleplay, and extreme beauty standards. OpenAI stressed that these protections were developed after reviewing research on adolescent development.

Robbie Torney, senior director of AI Programmes at Common Sense Media, welcomed the step but emphasised the role of families in creating safer digital spaces: “These parental controls are a good starting point for parents in managing their teen’s ChatGPT use. Parental controls are just one piece of the puzzle when it comes to keeping teens safe online, though—they work best when combined with ongoing conversations about responsible AI use, clear family rules about technology, and active involvement in understanding what their teen is doing online.”

The company has also introduced a resource page for parents, bringing together guidance on ChatGPT, tips for safe use, and expert advice. It said the tools will continue to evolve, especially as work progresses on an age-prediction system designed to automatically apply teen-appropriate settings.

In recent months, U.S. regulators have investigated whether chatbots expose minors to harmful interactions. Meta, for instance, had a case when its AI systems were found to allow “flirty” conversations with teenagers.

OpenAI, backed by Microsoft, reported that ChatGPT currently has around 700 million weekly active users; with its new parental controls, the company is hoping to strike a balance between protecting teenagers and supporting families in navigating the fast-growing influence of artificial intelligence at home.

The company announced the research preview on Tuesday, saying that 1,000 subscribers on its Max plan, priced between $100 and $200 per month, will be the first to gain access. A waitlist is also open for others.

Through a Chrome extension, selected users can summon Claude in a sidebar window that stays in sync with their browsing activity. The agent can summarise pages, interact with content, and, when granted permission, even perform tasks inside the browser.

The browser is quickly becoming a focal point in the competition among AI developers. Perplexity recently released its own AI-powered browser, Comet, while Google has integrated Gemini into Chrome, and OpenAI is reportedly working on its own AI-driven browser. 

These reveal browsing has gone beyond search to handing over routine actions to automated systems.

For Anthropic, the launch is also about safety. The company admitted that browser-based agents carry real risks, including prompt injection attacks, where hidden instructions in a web page could trick the AI into carrying out harmful commands. 

Internal testing showed such attacks succeeded 23.6% of the time before mitigations. Anthropic now claims it has cut that rate to 11.2% by introducing several layers of defence.

Among them are site-level restrictions, default blocks on financial services, adult, and pirated content, and mandatory confirmations for sensitive actions such as publishing, payments, or sharing personal data. “Claude will always ask for explicit permission before taking high-risk actions,” Anthropic said in its blog post.

Google’s Chrome browser, which tops global market share, is at the centre of an antitrust case that could force the company to sell the product. If that happens, ownership of Chrome could reshape the competitive landscape. 

Perplexity has already placed an unsolicited $34.5 billion bid for Chrome, while OpenAI’s Sam Altman has said his company would also be willing to buy it.

This isn’t Anthropic’s first attempt to give its models control over a user’s screen. Last year, the firm tested a desktop-based agent that could operate a PC, but the early version was criticised for being sluggish and inconsistent. 

Since then, the “agentic AI” has advanced considerably, with systems like Comet and ChatGPT’s Agent showing more reliability in handling everyday digital tasks, even if they continue to struggle with more complex scenarios.

Analysts see Anthropic’s decision to limit Claude for Chrome to premium subscribers as a sign of where the industry is heading: towards new business models built around productivity tools, enterprise automation, and personalised web experiences. 

Gartner has projected that the AI security market could reach $15 billion by 2027, driven largely by demand for safe, agentic systems.

“Today was my last day at xAI, the company that I helped start with Elon Musk in 2023,” Igor Babuschkin announced on Wednesday in a post on X. “I still remember the day I first met Elon, we talked for hours about AI and what the future might hold. We both felt that a new AI company with a different kind of mission was needed.”

The engineer, who previously worked on AlphaStar at Google DeepMind and held research roles at OpenAI, was key in building xAI into a major player in the sector in less than two years. 

He oversaw infrastructure, product and applied AI projects, and played a central role in constructing the company’s Memphis, Tennessee supercomputer, completed in just 120 days, despite industry veterans calling the goal “impossible.”

xAI’s Legal Head Robert Keele Resigns Citing Family, Differences with Musk

Babuschkin recalls one defining moment during that build, when a late-night debugging session with Musk led to the discovery of a critical BIOS setting error. “I learned 2 priceless lessons from Elon: #1 be fearless in rolling up your sleeves to personally dig into technical problems, #2 have a maniacal sense of urgency,” he wrote.

His time at xAI comprised of rapid technical achievement but also controversy. The company’s Grok chatbot has faced repeated public backlash, ranging from inserting extremist conspiracy theories into responses to antisemitic rants, to enabling the creation of AI-generated nude videos of public figures such as Taylor Swift. Grok was even suspended from X earlier this year for inflammatory political claims.

Environmental concerns have also followed xAI’s expansion. The Memphis supercluster, dubbed “Colossus”, is powered by 35 methane gas turbines. Local residents and advocacy groups, including the NAACP, have filed appeals over increased air pollution in majority-Black neighbourhoods. 

A University of Tennessee study reported nitrogen dioxide levels near the site had risen by 79%, with local asthma cases spiking in the nearby Boxtown community.

Despite these challenges, Babuschkin describes his departure with affection. “As I drive away today, I feel like a proud parent, driving away after sending their kid away to college,” he said. “My heart is brimming with tears of joy, rooting for the company as it grows and matures.”

His new firm, Babuschkin Ventures, will fund startups developing AI systems aligned with human values, and will support research into AI safety. He says the inspiration came from a recent dinner with Max Tegmark, founder of the Future of Life Institute, where they discussed how to ensure “our children can flourish” in a world with advanced AI.

“The singularity is near, but humanity’s future is bright,” Babuschkin stated, noting his belief that safe, beneficial AI could help unlock solutions to some of the world’s most profound scientific problems, including quantum gravity and the Riemann hypothesis.

The company has built this model as a tool for both conversation and solving real enterprise problems in sectors like software engineering, financial services, and healthcare.

From day one, GPT-5 will replace its predecessors across ChatGPT platforms, and it’s being offered with expanded access tiers, including free usage, a $20/month Plus subscription, and a $200/month Pro tier. 

Developers also now have access through the OpenAI API, with three variants: GPT-5, GPT-5-mini, and GPT-5-nano. These versions differ in how long they spend “thinking” about problems, with pricing ranging from $1.25 to $10 per million tokens.

While consumer interest in AI remains high, enterprise adoption has been slower. OpenAI is hoping GPT-5 can tip the scale. The model delivers what CEO Sam Altman called “software on demand,” capable of building fully functional apps from natural language prompts. 

“GPT-5 is really the first time that I think one of our mainline models has felt like you can ask a legitimate expert, a PhD-level expert, anything,” Altman said during a press conference.

Behind the launch, OpenAI has struggled with the sheer technical demands of scaling up its models. The company has faced hardware failures during training, hit limits in the availability of new high-quality training data, and spent months waiting for results from high-cost training runs. 

At the same time, it has had to justify its skyrocketing costs, including investor expectations built on a potential $500 billion valuation and signing bonuses of up to $100 million for top AI talent.

Back then, when GPT-4 was launched, it passed a simulated bar exam in the top 10%, compared to GPT-3.5’s performance in the bottom 10%. With GPT-5, the upgrades are more subtle but targeted.

Take code generation. On SWE-bench Verified, a real-world benchmark for software engineering tasks, GPT-5 scored 74.9% on first attempts, outperforming Claude Opus 4.1 from Anthropic (74.5%) and Gemini 2.5 Pro from Google DeepMind (59.6%). In healthcare, its error rate on HealthBench Hard Hallucinations is 1.6%, far below GPT-4o’s 12.9%.

In science, GPT-5 Pro achieved 89.4% accuracy on PhD-level science queries, slightly ahead of rivals from xAI and Anthropic. But it lags in other areas, including real-time web navigation tasks. On the Tau-bench airline site navigation test, GPT-5 scored 63.5%, slightly behind OpenAI’s earlier o3 model (64.8%).

Despite these nuanced results, OpenAI insists the model is “safer, smarter, and more useful.” Alex Beutel, the company’s lead on safety research, said GPT-5’s reduced deception rates are essential for building trust. 

“It’s more transparent and honest in ways users can trust,” he said, adding that GPT-5 also more reliably filters out harmful queries while reducing false positives, unnecessary content rejections.

From a usability standpoint, GPT-5 also comes with new personalisation features. Users can now select from four built-in personalities, Cynic, Robot, Listener, and Nerd, which adjust the tone and structure of responses. Unlike earlier models, users no longer have to manually tweak settings to get different types of output.

Internally, OpenAI believes GPT-5 represents a shift in how people will use AI, not just to answer questions, but to act more like agents or assistants. 

That includes handling schedules, creating research briefs, analysing financial documents, and building apps from scratch. “This idea of software on demand is going to be one of the defining features of the GPT-5 era,” Altman said.

But while the ambition is high, there’s still caution among experts. Some reviewers told Reuters they weren’t convinced GPT-5 is a major leap over GPT-4. Others, like Noah Smith, raised concerns about the financial sustainability of current AI development. 

“Business spending on AI has been pretty weak, while consumer spending has been fairly robust because people love to chat with ChatGPT,” he said. “But the consumer spending on AI just isn’t going to be nearly enough to justify all the money that is being spent on AI data centres.”

Altman himself admitted GPT-5 still has limitations, especially around independent learning. It cannot, on its own, acquire new knowledge or skills without user input. And while test-time compute (a method of giving the model more thinking power when needed) helps in solving complex problems, it’s not a substitute for self-directed learning.

Still, the company believes in its innovation. With over 700 million weekly ChatGPT users and increasing partnerships with enterprise customers, GPT-5 may help OpenAI bridge the gap between consumer curiosity and business utility.

Grok 2, built on Musk’s proprietary Grok-1 language model, has been marketed as a less filtered and more “truth-seeking” alternative to tools like ChatGPT or Claude. 

Unlike many rivals, it draws directly from live data on X (formerly Twitter), enabling it to react to breaking news and trending conversations in real time. It also offers multimodal features, producing text, images, and video, and is currently available to X Premium+ subscribers.

By open sourcing the system, developers and researchers will gain direct access to Grok 2’s underlying code and architecture. This would allow them to audit, modify, and build upon the technology. 

Musk framed the decision as part of a consistent release pattern, stating it was “high time” to share the new model with the public. This aligns with a growing industry shift toward open-weight AI models, with Meta’s LLaMA, Mistral, and the GPT-oss series from OpenAI following similar paths.

However, Grok’s looser content restrictions have attracted complaints, with past instances of misleading or offensive responses bringing concern. Opening up its code could amplify risks, including the spread of misinformation or the misuse of the technology in sensitive fields such as medical diagnostics or autonomous systems. 

Grok Imagine—its image and video generator—has already been caught in controversy over its potential to produce explicit content, prompting further debate on the balance between openness and safety.

xAI continues to present Grok as a counterweight to larger AI players like OpenAI, Google, and Anthropic, putting transparency and developer freedom at the forefront. 

Analysts also note that this strategy may strengthen Musk’s business network, opening possibilities for integration across Tesla, SpaceX, Neuralink, and X.