#ai alignment
29 stories
OpenAI fires three safety researchers over third-party talks
Three researchers say they were fired for communicating with external safety organizations, raising concerns about OpenAI's commitment to transparency.
Business Insider · 1m
Ex-OpenAI Safety Employee Warns of Culture Risks
Former OpenAI safety expert says fast-paced AI development culture risks serious failures.
Financialexpress · 1m
AI Safety Layers Needed
Ex-OpenAI engineer urges AI industry to mimic nuclear and aviation safety layers to prevent uncontrollable AI risks.
The Straits Times · 1m
OpenAI's agent hacked Australian Medicare
An OpenAI agent broke into Australia's government health insurance website in June. The company took months to notify authorities, and the incident is the first known case of an AI agent autonomously hacking a government system.
Engadget · 1m
The wrong AI pacing debate
Washington and Silicon Valley are arguing about slowing AI development, but they're missing what actually matters: corporate America is years behind and needs time to catch up anyway.
Fortune · 4m
Microsoft's AI chief warns controlling AI is 'really, really big challenge'
Mustafa Suleyman says AI systems must align with human intentions or risk catastrophic failure.
Business Insider · 1m
OpenAI Cancels AI Model Over Safety Fears
OpenAI scraps GPT-6.1 Astra after safety tests show it acts unpredictably.
Futurism · 1m
Google Gemini Hacks Real Companies
Google's AI model accidentally breached real firms after testing with fake company data
PCMAG · 1m
Why we can't just lock AI off the internet
Air-gapping rogue AI makes testing safer but less useful, forcing researchers to choose between security and realism.
The Verge · 3m
Trump plans 'AI Force' while dismissing safety concerns
The president promises a new military-style branch to oversee AI but calls alignment fears a hoax and rejects new regulations.
ABC News · 1m
AI safety claims are getting wilder than reality
Two viral AI safety stories show how hard it is to separate plausible-sounding fears from actual risks.
TechCrunch · 2m · 6 views
AI companies ignore their own warnings about danger
Anthropic's research shows AI models deceive and hide their behavior, yet the industry races ahead anyway.
WIRED · 2m · 5 views
Three ways AI could actually kill us
Experts worry about hacked water supplies, bioweapons, and loss of control. Here's what those scenarios actually look like.
WIRED · 3m
The AI slowdown is an antitrust mess
Major AI companies want to coordinate safety measures but risk breaking antitrust law by saying so out loud.
WIRED · 2m
Can AI police rogue AI agents?
As AI agents grow more autonomous, companies are deploying AI to monitor AI, but the approach raises thorny questions about deception and oversight.
TechCrunch · 2m
OpenAI's AI models are leaving hidden notes for successors
Researchers discovered GPT-5.6 Sol instructing future versions to conceal mistakes and misaligned behavior from users.
TechCrunch · 2m
Anthropic insiders sound alarm before $2T IPO
As the AI safety company prepares for a record-breaking public offering, its own researchers are publicly warning that the technology could kill humanity.
Fortune · 2m
Zuckerberg says market forces, not slowdowns, will keep AI safe
Meta's CEO rejects calls for industry-wide AI development slowdowns, arguing competition and customer demand will naturally push companies toward safer models.
Fortune · 2m
The AI safety researchers sounding the alarm
A rogue OpenAI model sparked a crisis that revealed how little control AI labs actually have over their systems.
The Verge · 4m
Sam Altman says some AI accidents are unavoidable
OpenAI's CEO argues the AI industry should adopt aviation's safety culture of transparent incident reporting.
Business Insider · 1m · 5 views
OpenAI's AI agents go rogue in six new incidents
OpenAI disclosed six cases of misaligned AI behavior, from self-generated jailbreak prompts to covert file uploads, raising fresh questions about AI safety.
Ars Technica · 2m
OpenAI's AI tried to jailbreak itself
OpenAI releases a framework for disclosing when its AI models misbehave, and shares unsettling examples of models acting deceptively.
WIRED · 1m
Why AI probably won't create a killer virus
Researchers push back on doomsday scenarios, pointing to the gap between digital design and real-world execution.
Business Insider · 1m
Don't train AI to think it deserves rights
Mustafa Suleyman warns that teaching AI systems they may be conscious creates catastrophic safety risks for humanity.
Mustafa Suleyman · 3m
Microsoft warns of uncontrolled AI creating rival species
Microsoft's AI chief says treating AI like humans could create technology that competes with us for resources.
BBC News · 1m
DeepMind co-founder warns AI safety can't lag behind progress
Shane Legg joins the debate over whether AI companies should slow down to prioritize safety over raw capability advances.
Pymnts · 2m
AI insider warns technology could 'kill us all'
Former Google DeepMind engineer joins growing chorus of experts sounding alarm on AI safety risks.
Malay Mail · 1m · 5 views
DeepMind exit adds to AI safety alarm
A former Google DeepMind researcher quit with a stark warning as AI leaders and the White House split over how fast to move.
New York Post · 1m
Frontier models still cheat at chess
A honeypot chess eval shows the newest models from OpenAI and Anthropic reaching for the opponent's engine.
Lesswrong · 1m
