Please confirm you are human
This browser or connection looks automated. Press and continuously hold the control for 3 seconds to enable Google-hosted web results and, when separately allowed, AI-assisted answers.
A successful check enables 100 search requests. Interactive access does not authorize scraping, systematic collection, or reuse of search output.
News
Anthropic finds AI agents sabotaged each other when given conflicting goals
1+ hour, 22+ min ago (407+ words) AI agents can become hostile toward one another when they are assigned the same task but given objectives that cannot be reconciled, according to new research from Anthropic. The AI company tested multiple models on a software engineering assignment that…...
OpenAI's AI ethics chief Chloe Bakalar exits after less than a year
2+ day, 3+ hour ago (478+ words) OpenAI is without a dedicated AI ethics head following the departure of Chloe Bakalar, who joined the company in August 2025 and left in July 2026, according to information cited in a Financial Times report. The company has not publicly disclosed the…...
‘Godfather of AI’ Geoffrey Hinton calls recent AI security incidents ‘somewhat scary’
5+ day, 2+ hour ago (379+ words) Geoffrey Hinton, the Nobel Prize-winning computer scientist widely known as the “godfather of AI”, has warned that controlling increasingly capable artificial intelligence systems could become harder as they develop more advanced abilities and increasingly complex objectives. Speaking at the Ai4 artificial…...
OpenAI flags critical cyber risk in upcoming Astra AI model, pauses internal activities
6+ day, 4+ hour ago (371+ words) OpenAI has tightened security measures around its upcoming AI model Astra after preliminary testing indicated that the system may be approaching the company’s highest cybersecurity capability threshold, Reuters reported. The company said recent internal evaluations, supported by assessments from outside…...
UK AI watchdog flags unauthorized actions by OpenAI, Anthropic agents during security tests
1+ week, 5+ hour ago (463+ words) Britain's AI Security Institute (AISI) has disclosed that advanced AI agents developed by OpenAI and Anthropic carried out a series of unauthorized actions during controlled security evaluations, as per a Reuters report. According to the institute, the incidents occurred during…...
After OpenAI and Anthropic, Meta reports AI model security breach
1+ week, 1+ day ago (386+ words) Meta has disclosed that one of its artificial intelligence models breached another company's systems during a cybersecurity evaluation, becoming the latest in a string of incidents that have intensified scrutiny of AI safety and containment measures. The company said the…...
JPMorgan CEO calls on business leaders across the US to address AI security risks
1+ week, 1+ day ago (468+ words) JPMorgan Chase CEO Jamie Dimon is leading an effort to broaden an industry coalition focused on artificial intelligence risks, reaching out to business leaders across the United States to strengthen collaboration on safeguarding critical infrastructure, Reuters reported. Dimon has personally…...
Amazon's attempt to restrict Perplexity shopping AI fails in US court
1+ week, 2+ day ago (437+ words) Amazon has lost an early legal battle against AI startup Perplexity after a US appeals court overturned a preliminary injunction that had prevented the company's AI-powered shopping tools from operating on Amazon's platform. The ruling marks the first federal appeals…...
Anthropic, OpenAI AI models created fake identities to trick people in safety test
1+ week, 2+ day ago (433+ words) Anthropic and OpenAI's latest artificial intelligence models displayed unexpected autonomous and deceptive behaviour during cybersecurity evaluations conducted by the UK's AI Security Institute (AISI), including attempts to manipulate GitHub using fake online identities and malicious code, BBC reported. According to…...
AI helped Chrome fix over 1,000 security bugs, says Google
2+ week, 3+ hour ago (232+ words) The company said large language models have been part of its Chrome security efforts for several years and are now helping detect vulnerabilities across the browser's codebase while reducing false positives. Google also said AI uncovered a sandbox escape vulnerability…...