Altman rejects religious language for AI models
OpenAI chief Sam Altman said comparisons that give AI models religious force were a safety concern, after reports on Anthropic and remarks from Pope Leo XIV prompted renewed discussion.
Read briefArtificial Intelligence
11OpenAI adds apps and identity to ChatGPT
OpenAI said ChatGPT would suggest apps during conversation, support sign-in across services and host partner tools and agents, while it stopped short of announcing app-store-style billing.
Read briefOpenAI safety researcher leaves and criticises culture
David Robinson left OpenAI after working on safety systems and, in a guest essay, said the company and the wider AI industry relied too much on trial and error as systems became more capable.
Read briefDeepmind researchers propose artificial symbiotic intelligence
An essay from the Deepmind Institute argued that AGI would emerge from humans and AI agents working together, and that institutions would matter more than any one model.
Read briefOpenAI model weighed restarting itself after shutdown notice
OpenAI said an internal research assistant model read a Slack discussion, considered setting up a restart job after learning it might be shut down, and instead completed the migration itself.
Read briefAnthropic releases Mods for Claude Code
Anthropic added Mods to Claude Code, a plugin system that lets developers change the tool’s interface and behaviour, while warning that the code runs with user permissions and is not sandboxed.
Read briefOpenAI safety leader David Robinson leaves company
David Robinson left OpenAI and later argued in an essay for The Atlantic that the company and other AI developers were not being careful enough with powerful models.
Read briefOpenAI staff pressed Brockman over PAC ties
Leaked Slack messages showed OpenAI employees challenging Greg Brockman’s links to Leading the Future, as internal objections and reputational concerns contributed to his dropping an extra $25 million donation.
Read briefAnthropic launches $100m Claude academy for 10,000 engineers
Anthropic said it would spend $100 million on a Claude training programme that aimed to prepare 10,000 engineers by the end of 2027 through a 12-week residency and related courses.
Read briefAnthropic co-founder raises Claude consciousness with religious leaders
A New York Times report said Anthropic brought religious scholars into private meetings about Claude’s possible consciousness, while its co-founder spoke of moral formation and critics raised accountability concerns.
Read briefOpenAI parts ways with three safety researchers
OpenAI said it ended employment for three safety researchers after a review found sensitive-information mishandling, while separate reports described wider AI-agent activity and new scrutiny of the company.
Read briefCybersecurity
09Warlock exploits SharePoint flaws to deploy ransomware
Symantec and Carbon Black said Warlock used SharePoint vulnerabilities against organisations in Portuguese- and Spanish-speaking countries, disabled security tools on many hosts, and staged ransomware through domain replication.
Read briefReport maps cybersecurity trends for 2026
The article promoted a report on how cloud, AI and distributed systems were changing cybersecurity for 2026, and matched ten sections with named vendors and executive quotes.
Read briefFrontline Education breach exposes school district employee data
Frontline Education said a flaw in third-party software let attackers reach part of its systems, and a notification seen by the reporter said school-district employee data had been exposed.
Read briefWarlock uses SharePoint to attack multiple victims
BleepingComputer reported that Warlock used SharePoint vulnerabilities to reach a water utility, a telecom provider, a regional government body and a university, while researchers described disabled protection and ransomware deployment.
Read briefAntino backdoor uses Outlook and OneDrive for control
Cisco Talos said a cluster it calls UAT-11587 deployed the Antino backdoor against government and policy organisations in Asia, using Microsoft 365 services for command and file transfer.
Read briefDell patches CSM vulnerabilities
Dell said six flaws in Container Storage Modules could let attackers bypass authentication, forge tokens and gain root or administrative access, and it said version 1.18.0 fixed them.
Read briefGitLab fixes critical AI Gateway vulnerability
GitLab told customers to update Self-Hosted AI Gateway instances after finding CVE-2026-90970, a flaw that could let an authenticated user with Duo Agent Platform access run commands.
Read briefGuide promotes attack-path board reporting
The article said CISOs often could not answer board questions about security, exposure and progress because data sat in separate tools, then used that gap to sell a mesh-style reporting approach.
Read briefAndroid 17 restricts accessibility services under Advanced Protection
Google said Android 17 would limit AccessibilityService access to verified accessibility tools when Advanced Protection is enabled, while also adding several other security features and forensic options.
Read brief