OpenAI said it resumed training its most advanced models after strengthening security and blocking internet access, and disclosed a previously unknown flaw in widely used software that it said had been fixed.
In Newport, Wales, a 23-metre-high hyperscale data centre has become a focus of tension between residents who object to its impact on their surroundings and politicians who present it as a source of jobs.
Session cookies, which keep users logged in, are a primary target for hackers. Stolen cookies can allow fraudsters to hijack accounts in a form of fraud known as session hijacking.
Cybersecurity professionals said AI is accelerating the discovery of software vulnerabilities and shortening the time attackers need to develop exploits, increasing the risk of disruptive attacks before systems can be patched.
Britain announced about £600 million, or $800 million, focused largely on defending against health and security threats from artificial intelligence. The package includes £100 million for an AI biosecurity laboratory and £65 million for pathogen countermeasures.
Governments and companies are racing to patch two PHP vulnerabilities discovered with help from AI models. Cybersecurity firms report a sharp rise in exploitation attempts, and the flaws could allow unauthenticated remote command execution on vulnerable web servers.
A TechCrunch analysis examines how AI-assisted vulnerability discovery and exploitation could expose weaknesses in foundational internet infrastructure. It says human researchers found vulnerabilities in the Hack the Planet initiative, while AI was used to exploit the bug.
OpenAI said it resumed training its most advanced AI models after tightening security controls and blocking internet access. It also disclosed a software vulnerability found while testing its defenses, amid concerns about AI agents identifying and exploiting flaws.
A political advertisement by Spain’s main opposition party used AI-manipulated likenesses of people linked to Prime Minister Pedro Sánchez’s Socialist party, including his wife, appearing to confess to wrongdoing, prompting outrage before next month’s election.
IT workers and cybersecurity experts told the BBC they had been working under intense pressure to update systems after vulnerabilities in PHP, reportedly uncovered with the help of AI, triggered a rare global cybersecurity response.
Researchers used AI to help identify a critical flaw in PHP, a programming language used across millions of websites. Cybersecurity leaders warned that AI could accelerate the discovery and exploitation of vulnerabilities in internet infrastructure.
OpenAI resumed reinforcement-learning training and internal use of its most advanced models after tightening security controls and blocking internet access in frontier-model training environments. It is also investigating vulnerabilities that could allow models to escape isolated systems.
Cybersecurity professionals told Reuters that advances in AI are uncovering a growing stream of software vulnerabilities. Researchers warned that AI-assisted discovery could lead to further disruption across critical internet infrastructure.
The UK chancellor is expected to announce £100 million in AI spending for a biosecurity laboratory described as a world first, as part of measures aimed at strengthening the country’s defences.
Microsoft researchers reported that an AI tool designed to diagnose patients sequentially was more than twice as accurate as general physicians on complex medical cases; the findings were published in Science.
Companies and governments rushed to patch critical PHP vulnerabilities after cloud firms reported a surge in hacking attempts, with some organizations taking websites offline as a precaution.
Researchers and security organizations warned of critical vulnerabilities in PHP, a programming language used by about 70% of websites worldwide. Organizations and governments have been urged to patch systems, with some temporarily taking websites offline.
A Chinese-speaking hacker reportedly used AI agents to autonomously probe and break into websites belonging to more than 100 companies, highlighting how artificial intelligence may accelerate offensive cyber operations with limited human involvement.
The proposed US Treasury rule would expand investment restrictions involving hypersonic systems, supercomputing, biotechnology and artificial intelligence. Treasury said the review reflects concerns that capital could accelerate military-relevant capabilities and technology transfer.
Researchers at cybersecurity startup VisionHeight identified 37 suspicious websites registered on the same night that appeared designed to influence AI-generated answers about elections, raising concerns about manipulated information reaching voters through chatbots.
Critical vulnerabilities in the widely used PHP programming language were discovered with the help of AI, prompting government agencies and other organizations across Europe to take websites offline as a precaution. Security experts warned that attackers could use AI to find and exploit vulnerable systems at scale.
Concerns about attacks exploiting PHP vulnerabilities led governments and companies to temporarily take websites offline while applying security patches. The flaws were reportedly found with help from OpenAI technology, raising questions about AI’s dual use in cybersecurity.
Google was ordered to stop work at two sites in Finland after an estimated 250,000 trees were cleared for a data centre. Residents said brown water from the site was entering a stream that feeds the Oulujoki River.
Technology companies are offering communities greater financial benefits and transparency as opposition grows to data centers supporting the artificial intelligence boom. Municipalities are passing moratoriums, while companies review practices such as nondisclosure agreements.
Ukraine said it still did not know why Starlink’s military satellite internet sister service, Starshield, was abruptly throttled, leaving troops vulnerable to Russian attacks and causing drones to fall from the skies.
Security researchers found a critical vulnerability in PHP, software used by as much as 70% of websites. The flaw was uncovered with help from artificial intelligence and could allow attackers to rapidly develop exploits.
Nvidia CEO Jensen Huang urged Europe’s artificial-intelligence industry to accelerate development and promoted modular systems designed to build AI data centers more quickly. He said Europe is moving from language models toward physical AI in manufacturing.
Meta’s earnings showed that increased spending on artificial intelligence had pushed the company’s cash flow into negative territory. The company said it would issue a new class of stock to raise more money.