Internet Research
- All
- News
- Videos
-
ChatGPT and Gemini Carry Gender, Race, Ethnic and Religious Biases, Claims Study
A new study from Pennsylvania State University has found that older versions of ChatGPT and Gemini were more prone to generating biased responses than other tested AI models. Researchers crowdsourced prompts designed to reveal bias and found reproducible results in 53 cases. The study identified eight bias types, including gender, race, and culture...
-
Lovable Partners With Guardio to Detect and Block Malicious Websites Created via Vibe Coding
Lovable has partnered with cybersecurity firm Guardio to add automated protection against malicious websites built on its AI coding platform. Guardio’s Safe Browsing Engine will scan every generated site for phishing, scams or impersonation and block suspicious projects before they go live. The move follows earlier research showing the platform c...
-
OpenAI Introduces Aardvark, an Agentic Security Researcher That Can Find and Fix Vulnerabilities
OpenAI has unveiled an artificial intelligence (AI) agent called Aardvark that can act as a software security researcher. Built on GPT-5, the agent can analyse code, detect vulnerabilities, and propose fixes. Currently in private beta, Aardvark is being tested by partner organisations. OpenAI says the tool aims to make software systems more secure ...
-
Tata Motors Reportedly Patched E-Dukaan, FleetEdge Flaws After Researcher Discovered AWS Key Leak
Tata Motors' two consumer-facing platforms, FleetEdge and E-Dukaan, were hacked in 2023 by a cybersecurity researcher, which were reportedly patched by the company in the same year. The flaws gave unauthorised access to bad actors to the platforms' Amazon Web Service (AWS) keys, allowing them to download Tata Motors' files, sensitive customer infor...
-
Anthropic Releases Microsoft Excel Support in Claude, Adds New Agentic Skills for Financial Tasks
Anthropic has introduced Claude for Excel, a new integration that lets users analyse and automate spreadsheet tasks directly through a Claude sidebar. The update also adds six finance-focused agentic skills and new data connectors for real-time access to sources like LSEG, Moody’s, and MT Newswires. These are currently available as research previ...
-
OpenAI Explains How It Assesses Mental Health Concerns of ChatGPT Users, Sparks Backlash
OpenAI has detailed its new safety evaluation system to detect signs of mental health distress, suicidal thoughts, and emotional dependence on ChatGPT. The company says the mechanism, built with input from clinicians and psychologists, helps its AI de-escalate sensitive chats and guide users to professional help. Critics, however, argue this approa...
-
Massive Data Breach Leaves 183 Million Email Credentials, Including Gmail Passwords, Exposed: Report
Cybersecurity researchers have discovered a massive data breach involving tens of millions of leaked Gmail accounts. According to a report, the breach occurred in April of this year and was recently made public, with a record of over 183 million credentials in total. It not only includes email accounts, but also the passwords that are associated wi...
-
‘Privacy Friendly’ Universe Browser Behaves Like Malware With Keylogging and Disabled Security Tools, Warn Researchers
Cybersecurity researchers have discovered a potential threat linked to a "privacy friendly" browser. Dubbed the Universe Browser, it is said to route its connections through servers in China and quietly installs several programs that covertly run in the background. The cybersecurity firm discovered that it was linked to a network of illegal gaming ...
-
Anthropic Launches Claude Code on the Web, Lets Users Assign Parallel Coding Tasks
Anthropic has expanded Claude Code to the web, letting paid users access the AI-powered coding tool without installing a CLI. Now available as a research preview, the web version supports GitHub integration and parallel coding tasks in isolated environments. The company has also launched a mobile version through its Claude iOS app.
-
Android Phones Susceptible to ‘Pixnapping’ Attack That Steals 2FA Codes, Messages and More, Researchers Say
Android devices are susceptible to a new type of attack that can covertly steal two-factor authentication (2FA) codes, private messages, location timelines, and other data. Security researchers have devised this new attack, called Pixnapping, which can exploit loopholes and security flaws in Android without requiring any system permissions. As per ...
-
OpenAI’s ChatGPT Pulse Will Deliver Daily Personalised Updates But at a Privacy Cost
OpenAI has introduced ChatGPT Pulse, a new AI feature that makes the chatbot proactive instead of purely reactive. Available only to top-tier subscribers, Pulse delivers daily personalised updates about events, meetings, or activities. It uses memory, chat history, feedback, and connected apps to curate content, although questions around data priva...
-
World’s First “All-Frequency” 6G Chipset Unveiled by Researchers
A group of researchers have published a paper in the Nature journal, highlighting the creation of a 6G chipset that can operate across a wide network spectrum between 0.5GHz and 115GHz. Capable of delivering more than 100Gbps mobile Internet speeds, the mobile platform was built on a new material dubbed thin-film lithium niobate (TFLN).
-
ChatGPT Provides Answers to Harmful Prompts When Tricked With Persuasion Tactics, Researchers Say
ChatGPT might be vulnerable to principles of persuasion, a group of researchers has claimed. During the experiment, the group used a range of prompts with different persuasion tactics such as flattery, peer pressure, and more, to GPT-4o mini, and found varying success rates. The experiment also highlights that breaking down the system hierarchy of ...
-
CloudSEK Research Reveals How AI Summarising Tools Can Be Tricked Using Prompt Injection-Based Attacks
CloudSEK, a cybersecurity firm, highlighted that artificial intelligence (AI) summarising tools can be tricked into carrying out commands of threat actors using benign CSS tricks. These tricks usually involve using hidden text in emails, messages, weblinks, and web pages. When a user asks an AI chatbot or an AI summarising tool to process the conte...
-
OpenAI Learning Accelerator Introduced in India, Students and Teachers to Get 5 Lakh ChatGPT Licenses
OpenAI Learning Accelerator, a new India-focused initiative, was introduced by the company on Monday. Through this initiative, the San Francisco-based artificial intelligence (AI) firm aims to build advanced AI research, training, and deployment capabilities among the educators and learners of the country. The AI firm is also collaborating with var...
-
ChatGPT and Gemini Carry Gender, Race, Ethnic and Religious Biases, Claims Study
A new study from Pennsylvania State University has found that older versions of ChatGPT and Gemini were more prone to generating biased responses than other tested AI models. Researchers crowdsourced prompts designed to reveal bias and found reproducible results in 53 cases. The study identified eight bias types, including gender, race, and culture...
-
Lovable Partners With Guardio to Detect and Block Malicious Websites Created via Vibe Coding
Lovable has partnered with cybersecurity firm Guardio to add automated protection against malicious websites built on its AI coding platform. Guardio’s Safe Browsing Engine will scan every generated site for phishing, scams or impersonation and block suspicious projects before they go live. The move follows earlier research showing the platform c...
-
OpenAI Introduces Aardvark, an Agentic Security Researcher That Can Find and Fix Vulnerabilities
OpenAI has unveiled an artificial intelligence (AI) agent called Aardvark that can act as a software security researcher. Built on GPT-5, the agent can analyse code, detect vulnerabilities, and propose fixes. Currently in private beta, Aardvark is being tested by partner organisations. OpenAI says the tool aims to make software systems more secure ...
-
Tata Motors Reportedly Patched E-Dukaan, FleetEdge Flaws After Researcher Discovered AWS Key Leak
Tata Motors' two consumer-facing platforms, FleetEdge and E-Dukaan, were hacked in 2023 by a cybersecurity researcher, which were reportedly patched by the company in the same year. The flaws gave unauthorised access to bad actors to the platforms' Amazon Web Service (AWS) keys, allowing them to download Tata Motors' files, sensitive customer infor...
-
Anthropic Releases Microsoft Excel Support in Claude, Adds New Agentic Skills for Financial Tasks
Anthropic has introduced Claude for Excel, a new integration that lets users analyse and automate spreadsheet tasks directly through a Claude sidebar. The update also adds six finance-focused agentic skills and new data connectors for real-time access to sources like LSEG, Moody’s, and MT Newswires. These are currently available as research previ...
-
OpenAI Explains How It Assesses Mental Health Concerns of ChatGPT Users, Sparks Backlash
OpenAI has detailed its new safety evaluation system to detect signs of mental health distress, suicidal thoughts, and emotional dependence on ChatGPT. The company says the mechanism, built with input from clinicians and psychologists, helps its AI de-escalate sensitive chats and guide users to professional help. Critics, however, argue this approa...
-
Massive Data Breach Leaves 183 Million Email Credentials, Including Gmail Passwords, Exposed: Report
Cybersecurity researchers have discovered a massive data breach involving tens of millions of leaked Gmail accounts. According to a report, the breach occurred in April of this year and was recently made public, with a record of over 183 million credentials in total. It not only includes email accounts, but also the passwords that are associated wi...
-
‘Privacy Friendly’ Universe Browser Behaves Like Malware With Keylogging and Disabled Security Tools, Warn Researchers
Cybersecurity researchers have discovered a potential threat linked to a "privacy friendly" browser. Dubbed the Universe Browser, it is said to route its connections through servers in China and quietly installs several programs that covertly run in the background. The cybersecurity firm discovered that it was linked to a network of illegal gaming ...
-
Anthropic Launches Claude Code on the Web, Lets Users Assign Parallel Coding Tasks
Anthropic has expanded Claude Code to the web, letting paid users access the AI-powered coding tool without installing a CLI. Now available as a research preview, the web version supports GitHub integration and parallel coding tasks in isolated environments. The company has also launched a mobile version through its Claude iOS app.
-
Android Phones Susceptible to ‘Pixnapping’ Attack That Steals 2FA Codes, Messages and More, Researchers Say
Android devices are susceptible to a new type of attack that can covertly steal two-factor authentication (2FA) codes, private messages, location timelines, and other data. Security researchers have devised this new attack, called Pixnapping, which can exploit loopholes and security flaws in Android without requiring any system permissions. As per ...
-
OpenAI’s ChatGPT Pulse Will Deliver Daily Personalised Updates But at a Privacy Cost
OpenAI has introduced ChatGPT Pulse, a new AI feature that makes the chatbot proactive instead of purely reactive. Available only to top-tier subscribers, Pulse delivers daily personalised updates about events, meetings, or activities. It uses memory, chat history, feedback, and connected apps to curate content, although questions around data priva...
-
World’s First “All-Frequency” 6G Chipset Unveiled by Researchers
A group of researchers have published a paper in the Nature journal, highlighting the creation of a 6G chipset that can operate across a wide network spectrum between 0.5GHz and 115GHz. Capable of delivering more than 100Gbps mobile Internet speeds, the mobile platform was built on a new material dubbed thin-film lithium niobate (TFLN).
-
ChatGPT Provides Answers to Harmful Prompts When Tricked With Persuasion Tactics, Researchers Say
ChatGPT might be vulnerable to principles of persuasion, a group of researchers has claimed. During the experiment, the group used a range of prompts with different persuasion tactics such as flattery, peer pressure, and more, to GPT-4o mini, and found varying success rates. The experiment also highlights that breaking down the system hierarchy of ...
-
CloudSEK Research Reveals How AI Summarising Tools Can Be Tricked Using Prompt Injection-Based Attacks
CloudSEK, a cybersecurity firm, highlighted that artificial intelligence (AI) summarising tools can be tricked into carrying out commands of threat actors using benign CSS tricks. These tricks usually involve using hidden text in emails, messages, weblinks, and web pages. When a user asks an AI chatbot or an AI summarising tool to process the conte...
-
OpenAI Learning Accelerator Introduced in India, Students and Teachers to Get 5 Lakh ChatGPT Licenses
OpenAI Learning Accelerator, a new India-focused initiative, was introduced by the company on Monday. Through this initiative, the San Francisco-based artificial intelligence (AI) firm aims to build advanced AI research, training, and deployment capabilities among the educators and learners of the country. The AI firm is also collaborating with var...
