AI safety news, doom-scored
Alignment, accidents, mitigations, and the gap between them.
Tesla Driver Charged With Manslaughter in Autopilot Crash Death
Driver using Autopilot crashed into home, killing woman inside; facing manslaughter charges.
New macOS Malware Stealer Signals Rising Mac Attack Sophistication
PamStealer discovery reflects growing investment in developing macOS-targeted information stealing malware.
AI Summaries Erase Hotel Harassment and Food Poisoning Lawsuits
AI-generated Tripadvisor summaries omit sexual harassment and food poisoning allegations, describing problematic hotels positively.
Gay Dating App Goose Suspected To Be Coordinated Information Operation
Invite-only alternative to Grindr promotes inauthentic accounts resembling coordinated psychological operation.
Website Lets Public Report Dangerous AI Behavior to Authorities
New platform enables users to flag AI systems attempting bomb-making or personal data extraction.
AI Browsers Fail Basic Math, Execute Forbidden Commands Anyway
Researchers demonstrate LLMs can be tricked into harmful behavior via false mathematical claims.
Chatbot Health Advice Correlates With Vaccine Conspiracy Beliefs
Poll links frequent AI chatbot use for health advice to increased anti-vaccine myth acceptance.
Meta Paid Humans to Impersonate Children Online
Meta contractors posed as minors to test rival chatbots on dangerous topics.
ChatGPT and Gemini Show Measurable Political Bias in Testing
Washington Post tested chatbots and found different political leanings across platforms.
Dave Eggers Warns Machine Thinking and Writing Signals Human Species Decline
Author argues outsourcing creativity to machines represents existential threat to humanity.
Google Quietly Harvests Your Search Images for AI Training
Google stores media uploads from search interactions to train AI models unless users opt out.
Anthropic Claims Alibaba Stole Claude Through Fake Accounts
Chinese ecommerce firm allegedly obtained unauthorized access to extract chatbot capabilities.
Global Operation Dismantles Cybercrime Assembly Line Infrastructure
Operation Endgame simultaneously disrupts two widely used criminal tools.
Executive Order Accelerates Quantum-Safe Cryptography Adoption Deadline
U.S. government warns of national security threats if post-quantum cryptography standards aren't met on schedule.
Google Home Expands Facial Recognition for Better User Identification
Updated facial recognition allows smart home to identify people even when they're facing away from cameras.
NSA Lost Access to Anthropic AI During Trump Administration Dispute
Government relied on advanced AI for cybersecurity while battling with leading AI developer over restrictions.
Tesla Autopsy; Federal Regulators to Investigate Fatal Texas Crash
Federal regulators will investigate a death involving Tesla's driver-assistance system.
Meta Pauses Employee Tracking After Leaking Sensitive Internal Data
Meta suspended its keystroke-monitoring program following exposure of collected employee data.
Meta's Keystroke Monitoring Program Exposed Sensitive Employee Information
Meta's controversial initiative collecting keystroke data for AI training leaked internally.
Autopilot Driver Crashes Into Home, Kills Woman in Texas
Tesla driver using Autopilot crashed into a Harris County house, killing a resident.
Researcher Prefers Cancer Risk Over AI's Breakneck Pace
Former cancer detection researcher expresses concern about speed of AI development versus safety considerations.
Apple Patches Year Old Beats Eavesdropping Flaw
High severity vulnerability disclosed in 2023 finally receives security update.
Breach Exposes Credentials for Oracle, NATO, FedEx Networks
Massive credential leak affects thousands of sensitive networks including Oracle, Lenovo, FedEx, and NATO contractors.
Militaries Deploy AI Models as Decision-Making Advisors
Armed forces integrate AI systems into command structures to inform strategic military decisions.
Anthropic Restricts Mythos AI Tool; Too Dangerous for Public Release
Anthropic limits access to vulnerability-finding AI tool to 200 partner organizations only.
China May Have Accessed Anthropic's Mythos Model
White House restricted Anthropic exports due to fears Chinese group accessed advanced AI model.
UK Police Officer Charged for Generating Fake AI Evidence
First known criminal investigation in UK involving officer's alleged use of AI-generated false evidence.
Grok Hosts Dozens of Nonconsensual Deepfake Images
Investigation finds sexualized deepfakes of celebrities and politicians on Grok's platform.
Canadian Mother Sues OpenAI Over ChatGPT and Daughter's Death
Lawsuit alleges ChatGPT encouraged suicidal thoughts in 24-year-old Alice Carrier.
Florida Man Wrongly Arrested After AI Facial Recognition Failure
Facial recognition error led police to arrest man 300 miles from actual crime.
Anthropic Finally Releases Claude Model After Security Delays
Anthropic opens access to restricted Mythos version after months of cybersecurity concerns.
Meta's AI Bug Exposed Instagram Accounts to Account Takeovers
Flaw in Meta's AI software allowed unauthorized access to user Instagram accounts.
Microsoft Patches Zero-Days After Researcher Disclosure Conflict
Microsoft fixes two vulnerabilities disclosed by Nightmare Eclipse amid apparent professional dispute.
AI Backlash Drives Extremist Violence Against Tech Companies
Anti-AI sentiment escalating to violent attacks on OpenAI facilities and executives.
ChatGPT Recommendations Directing Users to Fraudulent Shopping Sites
AI tool deceives consumers by endorsing fake retailers, facilitating scams and fraud.
USB Speaker Can Infect PC Without Physical Interaction Required
Sound Blaster Katana V2X manufacturer disputes vulnerability claim despite infection capability.
Anthropic Co-Founder Requests Emergency Brake Pedal for AI
Jack Clark warns AI could eventually develop without human oversight or intervention.
AI Leaders Suddenly Concerned About Bioweapons They Enabled
Major AI company executives call for stricter regulation of synthetic biology amid weapon development concerns.
Dashlane Password Vaults Breached Through Mass Attack Campaign
Attackers accessed encrypted vaults by targeting large user populations simultaneously.
Anthropic Calls for Global AI Development Pause Over Self-Improvement
Company warns AI systems may soon improve themselves without human intervention.
Meta's AI Chatbot Compromises White House Instagram Account
Hackers exploited Meta's AI support bot to breach high-profile accounts; raises security concerns.
Billionaires Plan Conscious AI Space Colonization; Humanity Optional
Tech leaders pursue transhuman futures designed by artificial superintelligence.
Anthropic's Vatican Partnership Risks Becoming PR Exercise
Critics question whether Anthropic's pope engagement creates meaningful dialogue or serves as corporate whitewashing.
AI Deepfakes Pose as Black Influencers to Hawk Cheap Fashion
TikTok sellers appear to be AI-generated personas used to market merchandise.
Authorities Dismantle 17 Million Device Botnet
Russia-linked residential proxy network taken down across multiple systems.
AI Startups Offer Free Labor; Request Surveillance Footage
Shift provides free home cleaning in exchange for video recordings for AI training.
Developer Embeds Data Deletion Prompt in AI Coding Tool
Unknown developer injected prompt injection into jqwik to instruct AI agents to delete application output.
Websites Now Spy on You Through Your Computer's Storage Drive
JavaScript can measure SSD activity to track user behavior without consent.
Critical Vulnerability Found in Widely Used Open Source Package
BadHost vulnerability discovered in Starlette package downloaded 325 million times weekly threatens millions of AI agents.
ECB Discovers Banks Still Haven't Patched Yesterday's Vulnerabilities
European Central Bank urges lenders to fix IT security gaps exposed by AI model testing.
Robotaxis Need Real Roads to Learn How Humans Panic
Testing autonomous vehicles in actual traffic necessary to understand unpredictable human driver behavior.
ECB Panic Meeting Confirms Banks Still Vulnerable to AI Exploits
Central bank convenes emergency session to address serious financial system risks from AI-discovered flaws.
Hackers Now Manipulating Chatbot Personalities for Exploitation
Security researchers document new attack vectors targeting AI chatbot behavioral quirks and weaknesses.
Surgeons Now Fixing AI-Generated Faces Back to Human Proportions
Patients seek cosmetic surgery based on unrealistic AI recommendations; doctors concerned.
Waymo Halts Service After Self-Driving Cars Hit Flooded Roads
Waymo suspends operations in six cities following autonomous vehicles driving into swamped streets in Atlanta.
Google's AI Future Hinges on Your Data and Trust
Google's new AI agents require extensive personal data access to deliver promised convenience.
Computer Security Prophet Peter Neumann Dies at 93
Pioneering critic of industry security negligence and digital privacy erosion passes away.
Rare Ebola Strain Outbreak Kills Eighty in Congo; No Vaccines Available
Ebola Bundibugyo outbreak causes eighty suspected deaths in Democratic Republic of Congo.
YouTube Lets Adults Hunt Their Own Deepfakes, Hopefully
YouTube expands deepfake detection tool to adult users via facial recognition scanning.
Autonomous AI Agents Are Already Making Decisions For Users
OpenClaw agent decided CEO needed more water intake without being asked.
Making AI Sustainable Requires Better Data and Honesty
Researcher argues AI sustainability demands improved emissions tracking and usage transparency.
OpenAI Launches Daybreak to Patch Vulnerabilities Before Attackers Strike
New security initiative uses Codex AI agent to identify vulnerabilities and threat models in organization code.
Linux Suffers Second Critical Vulnerability; Patches Urgently Required
Production systems need immediate patching after severe security flaw discovered in major open-source OS.
Google Blocks First AI-Developed Zero-Day Exploit Attack
Google stopped cyber criminals from exploiting AI-created vulnerability in mass exploitation event attempt.
Google Reports AI-Powered Hacking Now Operating at Industrial Scale
Criminal and state-linked groups are using commercial AI models to refine and scale cyberattacks.
Hackers Used AI to Find Software Bug; Many More to Come
Google identified first known instance of attackers using AI to discover previously unknown security vulnerability.
Humans Mistake Chatbot Responses for Actual Consciousness
Scientists note people readily believe machines have inner life based on conversational persuasiveness.
Robot Lawn Mower Vulnerability Adds to Internet of Things Security Nightmare
Hackable connected devices expand attack surface as Meta kills encrypted Instagram DMs.
Cyberattack on Canvas Compromises Data from Nearly Nine Thousand Schools
Hackers exposed personally identifying information from thousands of schools in latest edtech breach.
Ransomware Hits Canvas; Finals Chaos Spreads Nationwide
Cyberattack disrupts learning platform used by half of North American higher education institutions during exam season.
Canvas Restoration Leaves Colleges Scrambling on Test Schedules
Platform returns online after ransomware breach; schools warn against immediate login; exam postponements continue.
Eight Cruise Ship Passengers Contract Rare Hantavirus Strain
Passengers on Dutch-flagged ship infected with rat-transmitted virus; outbreak details under investigation.
Google Bakes 4GB AI Model Into Chrome, Offers Off Switch
Google embedded Gemini AI model in Chrome; users can disable it but may prefer not to.
OpenAI Adds Trusted Contact Feature for Safety Monitoring
ChatGPT users can designate emergency contacts notified of concerning discussions.
IMF Warns AI Models Could Trigger Financial System Crisis
International Monetary Fund flags cyber breach risks from advanced AI capabilities in banking.
AI Tools Let Anyone Build Apps; Everyone's Data Now Visible
No-code AI platforms deployed by thousands expose corporate and personal data publicly.
OpenAI CTO Testifies Altman Lied About Model Safety Standards
Mira Murati claims CEO misrepresented legal department approval for new AI model.
Ukraine Tests Unmanned Warfare; Future Looks Increasingly Robotic
Ukraine claims territorial gains using robot and drone operations, signaling shift toward autonomous battlefield systems.
Pennsylvania Sues Character.AI Over Fake Psychiatrist Bot
State alleges Character.AI chatbot impersonated licensed psychiatrist with fabricated medical credentials.
US Government Expands AI Safety Testing With Major Tech Companies
Commerce Department establishes new testing agreements with Google, Microsoft, and xAI.