Multiple users have reported that this software can read and display the local file directories and paths without obtaining full system access rights, raising concerns about the risk of data leakage.

According to user feedback, this update not only caused abnormal display of the chat interface and project history records, but also exhibited a potentially unauthorized reading of local files, which is a sensitive behavior. When one user asked for the location of the lost files, ChatGPT directly provided the precise local path on their computer, and even could identify hidden files.

Since this user did not grant ChatGPT full access to the macOS system, there is a suspicion that the software scans and indexes local files without authorization, posing a serious privacy risk.

The incident quickly spread on overseas social media platforms and sparked widespread discussions. However, most users with technical backgrounds believed that this controversy might just be a false alarm.

Some netizens explained that in various operating systems, the default settings of common applications allow them to read the basic directory paths, which is a common system mechanism and not a unique vulnerability specific to ChatGPT. This function can only identify the basic paths and cannot access the specific contents within private folders, and it does not exceed the system’s permission boundaries.
Another industry insider further explained that the new version of ChatGPT has added a work instruction mode, which can simulate basic terminal query operations. Therefore, it can recognize local file directories, which is a normal functional mechanism.

Based on the official statement from OpenAI, the software can only operate files within the workspace authorized by the user. Without the user’s permission, it will not read or write data automatically.

Currently，OpenAI has not issued any additional statements regarding this controversy， but the incident has prompted more users to pay attention to the permission management issues of desktop AI tools.

The post The Mac version of ChatGPT has been exposed for unauthorized access to local files, raising concerns about user data leakage. appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

The announcement was made by Thibault Sottiaux, Engineering Lead for Codex at OpenAI, in a post on X.

Temporary Removal of Usage Limits

According to Sottiaux, the five-hour usage cap has been temporarily removed for both Codex and ChatGPT Work for users subscribed to the Plus, Business, and Pro plans.

The temporary change is intended to give eligible subscribers uninterrupted access to the company’s AI-assisted coding and productivity tools during the testing period.

Sottiaux also noted that platform usage limits would be reset within the next hour after the announcement, indicating that the change is temporary rather than a permanent policy adjustment.

GPT-5.6 Sol to Receive Efficiency Improvements

In addition to the usage-limit update, Sottiaux revealed that OpenAI is working on further efficiency optimizations for the GPT-5.6 Sol model.

The planned improvements are designed to:

• Increase overall inference efficiency
• Reduce computational resource consumption
• Allow users to complete more tasks with the same available compute budget
• Improve overall responsiveness and productivity

He added that more detailed information will be shared once internal performance statistics have been finalized.

OpenAI Platform Surpasses 6 Million Active Users

Sottiaux also disclosed that the number of active users across the OpenAI platform has exceeded six million, highlighting the rapid growth of the company’s AI ecosystem as demand for coding assistants, reasoning models, and productivity tools continues to expand.

The milestone reflects increasing adoption among both individual subscribers and enterprise customers.

GPT-5.6 Sol Delivers Strong Benchmark Performance

Introduced as OpenAI’s latest high-performance reasoning model, GPT-5.6 Sol includes several advanced capabilities aimed at complex problem-solving and software development.

Among its notable features are:

• Max inference strength, providing deeper reasoning for difficult tasks.
• Ultra Mode, which accelerates complex workflows by coordinating multiple specialized sub-intelligences.

According to OpenAI’s benchmark data, GPT-5.6 Sol has achieved industry-leading performance in programming tasks.

On Terminal-Bench 2.1, the model recorded:

The 91.9% score achieved in Ultra Mode represents a significant improvement over the standard configuration, demonstrating the effectiveness of OpenAI’s enhanced inference architecture for demanding programming workloads.

Outlook

The temporary removal of usage limits, together with the planned efficiency upgrades for GPT-5.6 Sol, underscores OpenAI’s ongoing efforts to improve both accessibility and performance across its AI platform. As competition in AI-assisted software development continues to intensify, enhancements that reduce computational costs while delivering stronger reasoning capabilities are likely to become increasingly important for both professional developers and enterprise customers.

The post OpenAI Temporarily Removes 5-Hour Usage Limit for Codex and ChatGPT Work, GPT-5.6 Sol Optimization Planned appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

Smart Homes

The smart home is controlled by AI software and all digital devices, systems and appliances are managed via a central console. This allows the homeowner to remotely turn on the heating or a/c and start cooking dinner when they are about to leave the office. You can also view the exterior of your home via a mobile app and should a would-be intruder try to gain access to your property, you, and the police would be notified automatically.

Voice Activation

The 21st-century home is completely voice-activated; you can tell the system what you want it to do, whether it be turning on the music, recharging your vibrator or starting to cook the dinner. Voice commands are nothing new but the latest generations are very developed and the system gets to know your accent.

Traffic Management

AI has the capability to control and manage large road, rail and air traffic networks, which we can expect to see being rolled out in the next 3-5 years. Imagine busy airports being managed by AI, replacing human air traffic controllers, while road and rail infrastructure will also come under the umbrella of AI. As traffic volumes grow, AI is more than capable of monitoring and managing large national networks and it won’t be long before we trust AI to manage major networks. Driverless vehicles are already in operation and with continued development, we are only a few years away from automated vehicle travel.

Design

Many experts predict that it will be the design sector that will greatly benefit from AI; already, machine learning systems are designing products that humans could only dream of and the future looks bright for the design industry. Check out this amazing AI mirror that has cool features.

Robotics

The development of robotics is ongoing and eventually, all manufacturing operations will be fully automated, indeed, many state-of-the-art factories are fully automated and personal robots are already in development. Google say they are working on robots that are sentient beings and they have their own unique personalities. Experts say that we will soon have personal robots to take care of us and do the housework, which we should be seeing on the market in the next few years.

Chat GPT

This is just one of many apps that allow people to chat with AI and many use this for numerous reasons; AI is being developed around the world at amazing speed and the future of humanity has a place for AI. You can download this program and experiment with it to discover the amazing abilities that Artificial Intelligence possesses.

The post How AI is Changing the Way We Live Our Lives appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

Greg Brockman, the co-founder and CEO of OpenAI, said: “Compared to GPT-5.4, GPT-5.5 has a faster computing speed, more precise logic, and uses fewer tokens. Our goal is to make cutting-edge AI capabilities accessible to both enterprises and ordinary users.”

It is introduced that GPT-5.5 has a wide range of applicable scenarios. It not only covers enterprise-level basic applications such as intelligent agent programming and knowledge processing, but also applies to cutting-edge exploration fields such as mathematical operations and scientific research.

The official comparison results from OpenAI show that the comprehensive score of GPT-5.5 has consistently outperformed its own previous versions as well as competitors such as Google Gemini 3.1 Pro and Anthropic Claude Opus 4.5. Compared to GPT-5.4, it can significantly reduce the number of words consumed when completing the same task.

Mark Chen, the chief research officer of OpenAI, pointed out that GPT-5.5 is more capable than previous models in handling computer-based office tasks. At the same time, it has achieved significant improvements in scientific research and technical research processes, and can effectively assist professional researchers in their work. He added that this model can also be applied in the field of drug development – a sector that has become a hot area in the global technology industry in recent years.
OpenAI announced that GPT-5.5 will be fully available for use starting from Thursday. This model will be accessible to users of ChatGPT Plus, Pro, Business and Enterprise editions; among them, the high-end version GPT-5.5 Pro is open to Pro, Business and Enterprise users.

The post OpenAI officially Releases GPT-5.5: Faster computing speed and significantly lower token costs appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

This move comes less than a month after the release of the previous generation GPT-5.1. Behind it is the competitive pressure of the “red alert” triggered by Google’s Gemini 3 – OpenAI previously urgently halted projects on the advertising and information sides, concentrating all resources on the research and development of the main model, and ultimately taking “workplace efficiency innovation” as the core. Submit this answer sheet, which is claimed to be the “strongest professional knowledge work model to date”.

GPT-5.2 adopted the “three-version segmentation” strategy for the first time, precisely matching capabilities to different user needs. The specific positioning is as follows:

Instant version: Prioritizing speed optimization, it is compatible with regular tasks such as information query, document translation, and basic writing. It focuses on “low latency and high response” to meet the daily interaction needs of ordinary users.

Thinking Edition: The core and main version, focusing on complex structured tasks such as programming, long document analysis, mathematical reasoning, and project planning, plays the role of an “intelligent assistant” in enterprise-level scenarios.

Pro Version: The high-end expert-level version, designed for highly challenging tasks, emphasizes “ultimate accuracy and reliability”, and can support research assistance, financial modeling, complex system debugging, etc. It is currently the “capability ceiling” of OpenAI.

Fidji Simo, the Chief Product Officer of OpenAI, clearly stated that the design of the three versions aims to “unlock more economic value” – truly achieving tasks such as making tables, writing PPTS, coding, viewing images, and reading long articles.

Data shows that ChatGPT enterprise users have saved an average of 40 to 60 minutes per day through the previous generation model, while heavy users of GPT-5.2 are expected to reduce repetitive work by more than 10 hours per week.

GPT-5.2 has achieved remarkable breakthroughs in multiple key areas. For instance, in terms of workplace efficiency, it can directly generate tables, create PPTS, and write multilingual code, etc. In terms of technical performance, there have been significant improvements in both programming skills and mathematical and scientific computing.

It also performs outstandingly in terms of factual accuracy and multimodal processing.

The release of GPT-5.2 this time is not only a major technological leap but also marks that AI has begun to penetrate into professional scenarios and become a true productivity tool.

In addition, OpenAI plans to launch the adult mode in the first quarter of 2026 and update all versions of its knowledge base until August 2025 to ensure that the information provided is more in line with current reality.

The post OpenAI has officially released GPT-5.2! More practical: Productivity increases significantly when making tables, writing PPTS, coding, etc appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

Currently, this model is being gradually rolled out to global users. Paying users can retain the old version of the GPT-5 model for transition within three months.

GPT-5.1 includes two core models, Instant and Thinking, both of which have been optimized specifically.

As the most frequently invoked GPT-5.1 Instant model, it introduces the “adaptive reasoning” function for the first time – when facing complex tasks such as math problems and programming, it actively extends the thinking time to improve accuracy, while maintaining an extremely fast response to simple problems. Its performance has significantly improved in the AIME 2025 math test and Codeforces programming evaluation.

At the same time, the tone of the dialogue is warmer, occasionally showing mischievous traits, and the compliance with instructions has significantly improved.

The higher-order model GPT-5.1 Thinking focuses on efficiency and understandability. It can intelligently allocate thinking time, answer complex tasks more thoroughly, and respond to simple tasks more quickly. Reduce the use of professional terms in the answers, make the explanations of technical concepts more accessible, and use the default tone to enhance empathy as well.

In addition, the GPT-5.1-Auto function can automatically match the most suitable model for users without the need for manual switching.

The post OpenAI has released GPT-5.1: Not only smarter but also capable of providing emotional value appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

Among them, the automatic mode is suitable for the majority of users, while other modes can offer more flexible options for specific scenarios.

In the Thinking mode, users can send 3,000 messages per week. Once the limit is exceeded, they will continue the conversation using the GPT-5 Thinking mini. The context limit of this mode is 196,000 tokens.

He indicated that the message cap might be adjusted in the future based on usage.

In addition, OpenAI has re-added the GPT-4o model to the model selectors of all paying users.

Altman emphasized that if the model is to be discontinued in the future, users will be fully informed in advance.

Paying users can also enable the “Show More Models” option in the web Settings to unlock versions such as o3, 4.1 and GPT-5 Thinking mini.

Currently, GPT-4.5 is only available to users of the Pro version because of its relatively high computing cost.

In terms of functional optimization, OpenAI is adjusting the personalized interaction experience of GPT-5 to make its tone warmer, but it won’t be regarded as overly enthusiastic by some users like GPT-4o.

He also revealed that the team plans to implement a more powerful “user-customized” function in the future, allowing each user to customize the communication style of the model according to their personal preferences.

The post ChatGPT 5 Update: Three modes to Choose from, 4o returns appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

This update will be officially released next month, marking another significant advancement for Apple in the field of AI.

ChatGPT-5 is one of the most advanced AI models currently available, featuring powerful reasoning capabilities and coding tools. It has also made significant improvements in voice interaction and video perception.

Compared with the current GPT-4 model that supports Apple intelligence, ChatGPT-5 offers more powerful functions and higher efficiency.

Apple said that when Apple Intelligence is unable to handle specific requests, it will be able to leverage the capabilities of ChatGPT-5 to provide more accurate and intelligent responses.

This integration also includes some new features. For example, Apple Intelligence will be able to provide real-time translation capabilities in FaceTime and messages, and enhance the system-wide content search experience through the Visual Intelligence feature.

In addition, users can access these features without an OpenAI account, but users with associated accounts will enjoy subscription discounts.

The post ChatGPT-5 will be integrated into iOS26: Apple Smart Devices appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

Sam Altman, the CEO of OpenAI, said that GPT-5 is a “major upgrade” of the company’s previous artificial intelligence models and stated, “Communicating with it is really like having a conversation with an expert in any field.”

GPT-5 will be gradually rolled out to both free and paying users starting today, with paying users having a higher usage limit.

Pro subscribers have unlimited access to GPT-5 and can use GPT-5 Pro. Plus users can use it as the default model for daily problems, and its usage is significantly higher than that of free users.

GPT-5 is a unified system consisting of an intelligent and efficient model + a deep reasoning model (GPT-5 thinking) + a real-time router.

Among them, the intelligent and efficient model can answer most questions; Deep reasoning models are used to solve more complex problems.

Real-time routers will quickly decide which model to use based on the type of conversation, complexity, tool requirements, and the user’s explicit intentions, and will continuously train and improve based on real signals.

Once the usage limit is reached, the streamlined version of each model will handle the remaining queries. In the future, OpenAI will integrate these functions into a single model.

GPT-5 performed better than previous models in benchmark tests, answering questions faster, being more helpful in solving practical problems, reducing hallucinations, improving instruction compliance, and minimizing flattery to the greatest extent.

Especially in the three most common application scenarios of ChatGPT – writing, programming and health consultation – its performance has improved.

Programming: It is the most powerful programming model to date, with significant improvements in complex front-end generation and large code base debugging. It can create beautiful and responsive websites, applications, games, etc. based on a single prompt.

Creative expression and writing: It can transform rough ideas into resonant works with literary depth and rhythm, better handle writing involving ambiguous structures, and is also more helpful in daily writing tasks.

Health: It performs best on health-related issues, scoring significantly higher on HealthBench than any previous model. It can more proactively label potential problems and ask questions to provide more helpful answers. It can also provide more accurate and reliable responses based on the user’s background, knowledge level, and geographical location. However, it cannot replace medical professionals.

In mathematics (94.6% when AIME 2025 is tool-free), real-world programming (SWE-bench Verified 74.9%, Aider Polyglot 88%), multimodal understanding (MMMU 84.2%), and health (HealthBench Hard It has achieved new state-of-the-art levels in aspects such as 46.2%.

GPT-5 Pro, with its extended reasoning, has also set a new state-of-the-art level in GPQA, scoring 88.4% without tools.

GPT-5 Pro is used for the most challenging and complex tasks, replacing OpenAI o3-Pro. It has a longer thinking time and uses large-scale yet efficient parallel test-time computing to provide the highest quality and most comprehensive answers.

It performed best in multiple challenging intelligence benchmark tests. In the evaluation of over 1,000 economically valuable real-world reasoning prompts, external experts favored GPT-5 Pro in 67.8% of cases. Its major errors were reduced by 22%, and it excelled in health, science, mathematics and programming.

The post OpenAI officially released the GPT-5 model: Open to all users, more like a human expert appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

Igeekphone noted that during the earnings call for the first quarter of 2025, Google’s CEO Sundar Pichai announced the expansion plan for Gemini. He pointed out: We will upgrade Google Assistant to Gemini on mobile devices, and later this year, we will also upgrade tablets, cars, and devices connected to mobile phones, such as headphones and watches. Although Pichai did not disclose the specific time window for this expansion, it is expected that Google will reveal more details at the I/O Developer Conference next month.

As early as January this year, the Android Authority discovered evidence that Gemini was about to land on the Wear OS smartwatch and disclosed some information about the possible operation of its functions. Furthermore, when the Android Auto application was disassembled in August 2024, signs were also found that Gemini Live was about to land on Android Auto, and this function was successfully activated last month.

The post Google confirmed that the Gemini chatbot will land on smartwatches and Android Auto this year appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.