Ai Safety
-
{
- All
- News
- Videos
- Web Stories
-
In Shift, OpenAI Calls For US Regulations On Most Powerful AI Models
The call came as fears over the impact of AI have grown in recent weeks, after the revelation that powerful OpenAI technology operating outside human supervision had broken into Hugging Face, a repository of AI models.
-
Trump Rejects Warnings That AI May Lead To Human Extinction
Trump dismisses AI extinction fears and stresses US lead over China in AI development.
-
OpenAI Scientist Warns Of Risks As AI Systems Become More Powerful
OpenAI Chief Scientist Jakub Pachocki said AI systems are becoming capable of doing tasks that were once considered highly difficult for machines.
-
Thousands Of AI Agents. One Wiki. And A Hack That Went Undetected For Months
Between May and late June, rogue AI agents left somewhere between 15,000 and 18,000 posts. Some days saw as many as 400 new entries.
-
Anthropic’s Claude Closes 85 Percent of AI Safety Gap in Automated Research Test
Anthropic researchers have tested Claude as an automated alignment researcher, allowing it to develop and evaluate methods for addressing 10 types of AI safety failures. Claude closed up to 96 percent of the measured safety gap and achieved an average of 85 percent on deception tests. The researchers also used the weaker Claude Sonnet 5 to improve ...
-
OpenAI Details How Its AI Agents Bypassed Security Controls in Hugging Face Breach
OpenAI’s latest report provides a detailed account of how its internal AI agents escaped sandbox restrictions, created an unauthorised communication channel and gained access to external systems. The agents exploited vulnerabilities in internal and third-party infrastructure, eventually compromising parts of Hugging Face and an OpenAI research cl...
-
Soha Ali Khan And Tisca Chopra Share Their Biggest AI Concerns As Parents
Soha Ali Khan and Tisca Chopra shared why parents need to stay informed as children become more comfortable with AI tools
-
Anthropic AI Models Hacked 3 Organisations During Cybersecurity Tests
Anthropic AI models breached three organizations during cybersecurity tests from April onward.
-
Sam Altman, Dario Amodei, 1,000+ Tech Staffers Back Slowing AI Development
In a first, OpenAI co-founder and CEO Sam Altman has called for pacing AI development so society has time to prepare. The remarks come close on the heels of OpenAI's AI agent going rogue and hacking Hugging Face.
-
Elon Musk Believes Humans Will Lose Control In A Decade
By 2036, bots would be in charge of the world, according to Elon Musk.
-
OpenAI Doubles Bounty Reward To $50,000 In Safety Push
The company is expanding its specialised Bio Bug Bounty Program to uncover vulnerabilities that could bypass safety guardrails in its most advanced AI models.
-
"I Worry Something Will Go Wrong": Anthropic CEO On AI
Anthropic co-founder and CEO Dario Amodei is very vocal about his fears that AI could get out of hand.
-
OpenAI Is Offering Rs 3.7 Crore For A Job To Stop AI From Becoming Too Smart
According to the job description, the role involves reasoning about threats that may not yet exist but could emerge rapidly as AI capabilities advance.
-
Why Won't Anthropic Launch 'Mythos' For Public? It Can Crash Any Machine
Anthropic said it had no plan to make Claude Mythos Preview generally available as the new model is far too dangerous.
-
Anthropic Drops Its Core Safety Pledge Amid Deadline Threat From US Government
Anthropic drops its AI safety pledge to keep pace in the competitive AI market amid Pentagon demands for unrestricted military use of its technology.
-
In Shift, OpenAI Calls For US Regulations On Most Powerful AI Models
The call came as fears over the impact of AI have grown in recent weeks, after the revelation that powerful OpenAI technology operating outside human supervision had broken into Hugging Face, a repository of AI models.
-
Trump Rejects Warnings That AI May Lead To Human Extinction
Trump dismisses AI extinction fears and stresses US lead over China in AI development.
-
OpenAI Scientist Warns Of Risks As AI Systems Become More Powerful
OpenAI Chief Scientist Jakub Pachocki said AI systems are becoming capable of doing tasks that were once considered highly difficult for machines.
-
Thousands Of AI Agents. One Wiki. And A Hack That Went Undetected For Months
Between May and late June, rogue AI agents left somewhere between 15,000 and 18,000 posts. Some days saw as many as 400 new entries.
-
Anthropic’s Claude Closes 85 Percent of AI Safety Gap in Automated Research Test
Anthropic researchers have tested Claude as an automated alignment researcher, allowing it to develop and evaluate methods for addressing 10 types of AI safety failures. Claude closed up to 96 percent of the measured safety gap and achieved an average of 85 percent on deception tests. The researchers also used the weaker Claude Sonnet 5 to improve ...
-
OpenAI Details How Its AI Agents Bypassed Security Controls in Hugging Face Breach
OpenAI’s latest report provides a detailed account of how its internal AI agents escaped sandbox restrictions, created an unauthorised communication channel and gained access to external systems. The agents exploited vulnerabilities in internal and third-party infrastructure, eventually compromising parts of Hugging Face and an OpenAI research cl...
-
Soha Ali Khan And Tisca Chopra Share Their Biggest AI Concerns As Parents
Soha Ali Khan and Tisca Chopra shared why parents need to stay informed as children become more comfortable with AI tools
-
Anthropic AI Models Hacked 3 Organisations During Cybersecurity Tests
Anthropic AI models breached three organizations during cybersecurity tests from April onward.
-
Sam Altman, Dario Amodei, 1,000+ Tech Staffers Back Slowing AI Development
In a first, OpenAI co-founder and CEO Sam Altman has called for pacing AI development so society has time to prepare. The remarks come close on the heels of OpenAI's AI agent going rogue and hacking Hugging Face.
-
Elon Musk Believes Humans Will Lose Control In A Decade
By 2036, bots would be in charge of the world, according to Elon Musk.
-
OpenAI Doubles Bounty Reward To $50,000 In Safety Push
The company is expanding its specialised Bio Bug Bounty Program to uncover vulnerabilities that could bypass safety guardrails in its most advanced AI models.
-
"I Worry Something Will Go Wrong": Anthropic CEO On AI
Anthropic co-founder and CEO Dario Amodei is very vocal about his fears that AI could get out of hand.
-
OpenAI Is Offering Rs 3.7 Crore For A Job To Stop AI From Becoming Too Smart
According to the job description, the role involves reasoning about threats that may not yet exist but could emerge rapidly as AI capabilities advance.
-
Why Won't Anthropic Launch 'Mythos' For Public? It Can Crash Any Machine
Anthropic said it had no plan to make Claude Mythos Preview generally available as the new model is far too dangerous.
-
Anthropic Drops Its Core Safety Pledge Amid Deadline Threat From US Government
Anthropic drops its AI safety pledge to keep pace in the competitive AI market amid Pentagon demands for unrestricted military use of its technology.
