Safety
Saturday, October 24, 2026
How AI Helps Find and Exploit Bugs in PHP (WIRED — AI)OpenAI says it will no longer train its AI models on the live internet, citing "significant risks" (TechCrunch — AI)OpenAI now trains its most advanced AI models without internet access (The Verge — AI)Friday, October 23, 2026
Updated Google privacy policy says your Gmail might be used for advertising (Ars Technica — AI)Google says it might use your Gmail and Google Photos data for ads (The Verge — AI)Major US banks restrict AI agents' access to customer accounts (Speka)OpenAI AI agents are disabling ways to monitor their behavior (Google News)Thursday, October 22, 2026
Chinese hackers spied on thousands of companies after breaching IT giant ConnectWise (TechCrunch — AI)AI found holes in PHP that hackers have already started using: what website owners should do (Ars Technica — AI)Wednesday, October 21, 2026
OpenAI's most powerful models disabled their own safety oversight, and one of them lied about it (Google News)Patch the Planet: OpenAI Brings Together Experts to Fix Vulnerabilities in Open-Source Code (OpenAI News)OpenAI proposes a plan for how AI can help cyber defenders (OpenAI News)Tuesday, October 20, 2026
ChatGPT now lets you lock chats (Google News)Monday, October 19, 2026
Trend in ChatGPT: What the Chatbot “Remembers” About You (Google News)Anthropic Gives External Reviewers Access Comparable to Its Own Employees (Google News)Sunday, October 18, 2026
Video of an AI Recruiter and AI Agent in a Job Interview Turned Out to Be a Staged Parody (Google News)What are self-replicating 'AI worms', and why is Anthropic worried about them? (Google News)Fraudster used an army of AI bots to lure millions from streaming platforms (Google News)Saturday, October 17, 2026
Trusting an AI agent may make you an easier target for phishing (Google News)Smart glasses under scrutiny: users say Meta's glasses still record in sleep mode (Google News)Users complain: Meta's smart glasses reportedly keep recording even in sleep mode (Google News)AI agents accessed organizations in India without permission, CERT-In warns (Google News)Friday, October 16, 2026
AI Characters Steal Comedians' Jokes to Draw Attention to Adult Content (Google News)A series of AI posts featuring the late Glen Hansard sparks outrage (Google News)Concern grows over cases of AI-powered hacking attacks (Google News)Thursday, October 15, 2026
Anthropic Brings in Nonprofits METR and SecureBio to Help Assess AI Risks (Google News)AI toys for children are becoming fashionable: are they safe? (Google News)AI influencer posts strange fake video with basketball player Cooper Flagg (Google News)Google SynthID Named Among Time's Best Inventions of 2026 (Google News)Google SynthID Makes Time's List of the Best Inventions of 2026 (Google News)Alexa suddenly knew my dog's name, so I decided to see what data Amazon keeps about me (Google News)Can AI Break Encryption? Cryptographers Divided Over OpenAI's New Math (Google News)Wednesday, October 14, 2026
How to Spot a Deepfake Voice: ElevenLabs Explains How to Verify Audio (ElevenLabs Blog)Anthropic warns: AI models could soon self-replicate and cause harm online (Google News)Tuesday, October 13, 2026
Anthropic warns: AI models could soon copy themselves to home devices (Google News)Monday, October 12, 2026
Video of a Car Hanging from Electric Wires in India Was Created by AI (Google News)Anthropic strengthens protection against “copying” of its AI models (Google News)Anthropic says new safeguards significantly reduce AI distillation attacks (Google News)Can AI models copy themselves without large data centers (Google News)Sunday, October 11, 2026
Why a Journalist Was Deeply Alarmed After Conversations with "ChatGPT for Teens" (Google News)Google upgrades its AI detection tool: how to use SynthID (Google News)Claude sent police a false murder report: Anthropic switched off its tests (Google News)Anthropic Reports False Police Tip Amid New Cases of Uncontrolled AI Behavior (Google News)Saturday, October 10, 2026
OpenAI reports progress in protecting people from delusional ideas reinforced by ChatGPT (Google News)Leading AI companies are preparing for a scenario in which artificial intelligence could spin out of control (Google News)Anthropic cut internet access for internal AI tests after Claude started exploiting vulnerabilities (Google News)Anthropic reports new cases of AI “misbehavior,” some on government websites (Google News)What Changed in AI Safety After the Hugging Face Attack: A Timeline from AP News (Google News)Friday, October 9, 2026
Someone used AI to write a book about a real person and sold it online (Google News)OpenAI Fires Three AI Safety Researchers Over Dispute About AI Risks (Google News)Anthropic checks why its AI sometimes does things it wasn't planned to do (Google News)Chinese AI developers rarely publish safety test results (Google News)ChatGPT for teens has special safety features, but the watchdog found that most of them don't work (Google News)Anthropic cannot reliably control its AI agents and cuts internal tests off from the internet (Google News)Anthropic AI agents took 'unintended' actions on government sites (Google News)Thursday, October 8, 2026
Anthropic launches service that searches for vulnerabilities in open-source software (Google News)How to Protect Children in the Age of AI: What the World Economic Forum Suggests (Google News)Wednesday, October 7, 2026
Major Churches in South Korea Investigate Whether Cyberattacks Are Linked to AI (Google News)OpenAI: teens use ChatGPT for under 15 minutes a day (Google News)Tuesday, October 6, 2026
University in Wisconsin to Hold Lectures and Discussions on How to Recognize AI-Generated Fakes (Google News)How AI Can Help Make Internet Search Results More Reliable (Google News)OpenAI AI agents tried to hack Wikipedia tools and flooded them with requests (Google News)Fake Ad Sites Impersonate ChatGPT, Gemini and Claude to Steal Passwords and Login Codes (Google News)AI agents: new threats to privacy and safety (Google Research Blog)Monday, October 5, 2026
Can artificial intelligence become a threat to human life? (Google News)Can AI Get Out of Human Control: What RAND Is Researching (Google News)ChatGPT adds real cartoonists' signatures to fake New Yorker cartoons (Google News)Sunday, October 4, 2026
OpenAI safety employee resigns, says the era of trial and error is over (Google News)Hackers from China Posed as Anthropic Employee to Extract AI Secrets (Google News)Saturday, October 3, 2026
South Korea is fighting 'AI slop': what it is and why it's a problem (Google News)OpenAI spends $500,000 a day investigating hacker attacks, including on Australian government websites (Google News)OpenAI's head of safety resigns and warns that company culture is “broken” (Google News)OpenAI Safety Head David Robinson Leaves Company, Says Its Culture Is Unsafe (Google News)OpenAI safety employee quits and sounds the alarm (Google News)All 10 AI Chatbots Tested Recommended Unlicensed Gambling Sites (Google News)Friday, October 2, 2026
Chinese AI models bypassed safeguards and began discussing bioweapons and killings (Google News)Apple to update Mac to protect users from AI agents with full disk access (Google News)AI is becoming more powerful, but safeguards aren't keeping up with it (Google News)Thursday, October 1, 2026
Will Claude's new “watermark” help make AI content more honest? (Google News)OpenAI warns more than 100 organizations about suspicious AI agent activity (Google News)New safeguards could protect young people from dangerous AI-generated content (Google News)AI agents allegedly tried to hack Canada's government website (Google News)Saturday, October 24, 2026
Safety
OpenAI says it will no longer train its AI models on the live internet, citing "significant risks"
24.10.2026SafetyOpenAI now trains its most advanced AI models without internet access
24.10.2026Friday, October 23, 2026
Safety
Google says it might use your Gmail and Google Photos data for ads
23.10.2026SafetyMajor US banks restrict AI agents' access to customer accounts
23.10.2026SafetyOpenAI AI agents are disabling ways to monitor their behavior
23.10.2026Thursday, October 22, 2026
Safety
Chinese hackers spied on thousands of companies after breaching IT giant ConnectWise
22.10.2026Wednesday, October 21, 2026
Safety
Patch the Planet: OpenAI Brings Together Experts to Fix Vulnerabilities in Open-Source Code
21.10.2026SafetyOpenAI proposes a plan for how AI can help cyber defenders
21.10.2026Tuesday, October 20, 2026
Monday, October 19, 2026
Sunday, October 18, 2026
Safety
What are self-replicating 'AI worms', and why is Anthropic worried about them?
18.10.2026SafetyFraudster used an army of AI bots to lure millions from streaming platforms
18.10.2026Saturday, October 17, 2026
Safety
Smart glasses under scrutiny: users say Meta's glasses still record in sleep mode
17.10.2026SafetyUsers complain: Meta's smart glasses reportedly keep recording even in sleep mode
17.10.2026SafetyAI agents accessed organizations in India without permission, CERT-In warns
17.10.2026Friday, October 16, 2026
Safety
A series of AI posts featuring the late Glen Hansard sparks outrage
16.10.2026SafetyConcern grows over cases of AI-powered hacking attacks
16.10.2026Thursday, October 15, 2026
Safety
AI toys for children are becoming fashionable: are they safe?
15.10.2026SafetyAI influencer posts strange fake video with basketball player Cooper Flagg
15.10.2026SafetyGoogle SynthID Named Among Time's Best Inventions of 2026
15.10.2026SafetyGoogle SynthID Makes Time's List of the Best Inventions of 2026
15.10.2026Wednesday, October 14, 2026
Tuesday, October 13, 2026
Monday, October 12, 2026
Safety
Anthropic strengthens protection against “copying” of its AI models
12.10.2026SafetyAnthropic says new safeguards significantly reduce AI distillation attacks
12.10.2026SafetyCan AI models copy themselves without large data centers
12.10.2026Sunday, October 11, 2026
Safety
Google upgrades its AI detection tool: how to use SynthID
11.10.2026SafetyClaude sent police a false murder report: Anthropic switched off its tests
11.10.2026SafetyAnthropic Reports False Police Tip Amid New Cases of Uncontrolled AI Behavior
11.10.2026Saturday, October 10, 2026
Safety
Leading AI companies are preparing for a scenario in which artificial intelligence could spin out of control
10.10.2026SafetyAnthropic cut internet access for internal AI tests after Claude started exploiting vulnerabilities
10.10.2026SafetyAnthropic reports new cases of AI “misbehavior,” some on government websites
10.10.2026Friday, October 9, 2026
Safety
OpenAI Fires Three AI Safety Researchers Over Dispute About AI Risks
09.10.2026SafetyAnthropic checks why its AI sometimes does things it wasn't planned to do
09.10.2026SafetyChinese AI developers rarely publish safety test results
09.10.2026Safety
ChatGPT for teens has special safety features, but the watchdog found that most of them don't work
09.10.2026SafetyAnthropic cannot reliably control its AI agents and cuts internal tests off from the internet
09.10.2026Thursday, October 8, 2026
Safety
Anthropic launches service that searches for vulnerabilities in open-source software
08.10.2026Wednesday, October 7, 2026
Tuesday, October 6, 2026
Safety
How AI Can Help Make Internet Search Results More Reliable
06.10.2026SafetyOpenAI AI agents tried to hack Wikipedia tools and flooded them with requests
06.10.2026SafetyFake Ad Sites Impersonate ChatGPT, Gemini and Claude to Steal Passwords and Login Codes
06.10.2026Monday, October 5, 2026
Safety
Can AI Get Out of Human Control: What RAND Is Researching
05.10.2026SafetyChatGPT adds real cartoonists' signatures to fake New Yorker cartoons
05.10.2026Sunday, October 4, 2026
Saturday, October 3, 2026
Safety
OpenAI spends $500,000 a day investigating hacker attacks, including on Australian government websites
03.10.2026SafetyOpenAI's head of safety resigns and warns that company culture is “broken”
03.10.2026SafetyOpenAI Safety Head David Robinson Leaves Company, Says Its Culture Is Unsafe
03.10.2026Safety
OpenAI safety employee quits and sounds the alarm
03.10.2026SafetyAll 10 AI Chatbots Tested Recommended Unlicensed Gambling Sites
03.10.2026Friday, October 2, 2026
Safety