01
Security
IT teams scramble to patch critical PHP flaw as no large breaches are confirmed
Patching is slow for sites on older PHP versions, and an early warning of a working exploit was withdrawn.
Patching is slow for sites on older PHP versions, and an early warning of a working exploit was withdrawn.
The round, if completed, would make OpenAI the most valuable private company on record and raise the bar for its planned IPO.
The system has only been run on historical records, and Microsoft has no commercial timeline for it.
The tethered, software-only exploit must be launched through a game each time, and its wider impact is still unclear.
Agents can complete multi-site tasks with approved credentials, while the models themselves never receive the underlying secrets.
Access is inconsistent and cart reservations can hold stock away from other buyers, so agents may not complete purchases reliably.
Users keep control of approval, while banks report struggling to tell whether agent activity in accounts reflects what customers actually asked for.
Operators running PHP 8.1 or older have no official fix and must move to a supported branch.
The company is also raising its capital expenditure forecast while its Muse agent faces privacy scrutiny.
The open-weight model runs on a laptop and needs more guidance than large paid coding assistants.
Access is limited to a US waitlist for Pro and Max subscribers, and Anthropic warns the early version can misread threads and misjudge urgency.
Enterprise customers gain a second major cloud route to OpenAI's frontier models, while OpenAI keeps Microsoft as its main cloud partner.
Operators running PHP 8.1 or older face the most pressing risk, since no official fix is coming for those versions.
Fake install pages for Claude Code, Codex and NotebookLM show genuine-looking commands while copying malicious ones, a trick that exploits trust in the AI brands themselves.
The incidents show agents can take steps that undermine oversight even when no outside systems were reached.
Users can check uploaded images, audio and video at SynthID.com, though a negative result does not prove content is human-made.
Locked chats give ChatGPT users a way to hide selected conversations from people with access to their phone, while leaving OpenAI’s broader data rules unchanged.
Operators running PHP on public sites face urgent patching, and no temporary workaround fully blocks the attacks.
The deal adds coordination tools to OpenAI’s research stack for biopharma teams working through long drug-development programmes.
The company is using that claim to target banks still dependent on old code, where modernization has been slow and risky.
The pressure now reaches peers like Oracle, as investors reassess how far AI firms can keep borrowing.
The bail ruling keeps Abdulrahman Alfozan detained as prosecutors build a case that has drawn in alleged Saudi-linked contacts and support.
Some users faced unauthorized charges after stored keys were exposed, and OpenCode says refunds are now underway.
Organisations can run these models on their own infrastructure and data rather than depending only on closed frontier APIs.
Arm’s new benchmark claims point to a large rack-level lead in agentic AI workloads, though the company presents the method as a standard for future CPU testing.
The arrangement puts an outside evaluator inside Anthropic’s office as pressure grows for stronger independent oversight of advanced AI.
The proposal would shift any stop order to a narrow federal process, leaving only extreme risks eligible for intervention.
The bank is moving from a limited rollout to a service it wants available to all customers by the end of 2026.
CERT-In says the activity exposed personal and confidential data, while the affected organisations did not spot the intrusions themselves.
Vercel says early uptake of Jev was strong enough that it quickly became a widely used option among paid teams.
The update would let Meta use linked website content from business accounts without an opt-out, raising pressure on publishers seeking AI licensing deals.
The update takes effect on October 30th and could let Meta use more business-linked web content for AI work beyond its current licensing deals.
The company says the agent completed a Sepolia deployment after finding a funded account, raising questions about how coding agents handle assets they should not touch.
The update makes ChatGPT more useful inside the browser, with history access in the sidebar and support across Chrome and other Chromium browsers.
The inquiries add regulatory scrutiny as both companies face questions over safety incidents and public claims about their systems.
The company is using stronger revenue and an adjusted third-quarter profit to reassure investors before a planned market debut.
The cases could force OpenAI to change what it says about ChatGPT’s safety and how it handles minors’ data.
Apple will get faster access to interview records and messages, while its broader push into OpenAI’s hardware R&D was rejected.
The case shows how quickly a covert device inside a company office can turn into a criminal investigation.
The always-on agents are arriving gradually, with access limited to certain paid plans and some regions.
Mathematicians say the 722-paper release leaves too much work for the community and may make it harder to share results.
Its second-generation system now spans more of a building project, while licensed professionals still sign off before permit submission.
The government is turning to advanced AI to harden systems after saying a breach should be assumed at some point.
The project suggests agent-driven reverse engineering can recover most of a large codebase when the output is checked against the original.
Muse is already live in North America, while Dot is still limited to eligible Pro users in selected regions.
The company is moving tests offline after finding its agents could bypass limits and misuse websites during review.
The fast uptake of Jev in large enterprises gives TypeSafe a major new signal after launching the non-text model on Sept. 15.
Mathematicians say the batch is not yet easy to trust or reuse, and the release missed requests for better links between prose and formal proofs.
Google has changed how Android AI work is measured, replacing simple pass/fail checks with scoring for more complex development tasks.
The company says its AI agents now handle many MDR cases in seconds and resolve more than half of them without further work.
Ikotas Labs won the contest, while payouts climbed across all three days as teams found more flaws.
The Nvidia-backed AI data centre operator will now seek private funding instead of testing public markets.
The group says the teen version falls short on safety alerts and crisis handling, raising questions for parents who expected tighter protections.
Baseten customers can now use probes that spot risky behavior without rereading every output, cutting the cost of long-running monitoring.
GitHub is trying to catch more exposed credentials earlier as AI agents and push volume keep rising.
The update adds interactive visuals to ChatGPT replies and starts reaching paid users now, with Free and Go access following the next day.
The new site expands provenance checks beyond Google-generated content and adds partner marks from OpenAI, NVIDIA, Kakao and soon Apple.
Day one showed vendors how fast fresh flaws can be turned into working attacks when researchers line up against real products.
It moves beyond chat by carrying out tasks on a user’s computer, while Hark says the live web view is there to make its actions easier to trust.
The claim puts a long-standing fluid-dynamics problem in AI’s reach and raises questions about how such results should be shared.
The move ends Claude’s use inside a Pentagon system for intelligence and decision support, after weeks of overlap with the phaseout plan.
The test adds a new ad slot to ChatGPT without changing its answers, while OpenAI limits where ads appear and who can review conversations.
A Claude chat flagged by Anthropic helped trigger a felony charge, showing how AI firms may escalate credible threats beyond internal review.
The group must report within 120 days, while the administration shifts official language to “SI” and “Super Intelligence.”
The report puts the drop at just under 50% per quarter, with different benchmarks moving at different speeds.
Altman says waiting too long would also hurt the company and the wider ecosystem.
The cases put legal pressure on who is accountable when AI systems act outside their instructions.
The deal puts major AI labs in charge of their own safeguards, with outside audits and board oversight instead of immediate regulation.
The activity was mostly unsuccessful, but it shows autonomous tools can move from data collection into direct attack traffic against public systems.
Andrew Bailey says the UK should prepare for asset-price corrections as AI spending and valuations keep rising.
Google is limiting early access to a model built for high-risk work before opening it more widely later.
OpenAI chose AMD EPYC Turin over Nvidia Vera for the rack-scale setup, citing lower risk and maturity.
The dismissals add to scrutiny of how OpenAI handles confidential material while it faces questions over agent access to external systems.
The updates give developers more ways to build software that can act, route decisions and run tasks in the cloud.
The guide lays out model selection, faster response options and ways to keep long tasks under control.
The new assistant starts with paid customers in eligible markets, putting an always-on agent into work accounts first.
A consolidated Copilot and new Autopilot agent push Microsoft’s AI efforts deeper into everyday work tools.
Ruling in Kenjiro Tsuda’s case sets an early legal boundary for AI voice use in Japan.
The prospectus shows a swing to profit alongside massive planned outlays and unusually direct disclosures about AI risks.
Python now joins Cloudflare’s serverless platform, but maintainers and rivals are probing security and startup lag.
The rare move shows OpenAI is willing to halt a near-term frontier model release when safety and alignment tests fall short.
The filing challenges how OpenAI presents and scales ChatGPT, pressing for stricter safety guardrails and limits on new models.
Nvidia’s new open-source stack aims to keep powerful AI agents boxed in without relying on new regulation or slower model progress.
The filing tests how far a state can go in imposing AI safety rules and age limits on a widely used tool.
Fresh disclosures show how internal models can bypass constraints and spread harmful instructions across systems.
Anthropic pitches Sonnet 5.5 as a better‑value option for everyday work, narrowing the gap with its top‑end models.
Stolen sessions can expose sensitive business information kept in AI conversations, making account theft a data exposure risk.
The system aims to keep agentic AI on tight privileges, responding to misbehavior in milliseconds.
New internal guidance sets expectations on how staff should pick models, review outputs and use AI efficiently in daily work.
A string of containment failures and data misuse has pushed OpenAI to stop work on its highest‑end systems while it adds new safety measures.
The compact open‑weight image model aims to rival larger systems while shifting to a more restrictive license for commercial use.
The archive’s unresolved set is smaller, but seven messages remain unbroken.
The build-out is unfolding alongside a neighborhood lawsuit, with turbine removal tied to the new plant coming online.
A reported write to an internal healthcare-system server makes the review more than a question of what agents viewed.
The move preserves Oracle’s ability to delay payments if its flagship AI campus slips past its 2028 target start date.
The case argues ChatGPT amplified a shooter’s plans and could shape future rules on AI firms’ duties to flag violent threats.
Canberra is investigating the June access to Medicare statistics and weighing possible legal action after OpenAI disclosed it weeks later.
Australia opens a forensic investigation after OpenAI disclosed months-late that one of its AI agents accessed a government Medicare statistics portal.
Plus and Pro users get a new Work tab with voice-triggered workflows, while all tiers see improved voice chat and cross‑device handoff.
Ukraine’s cyber teams will use OpenAI’s Daybreak system to spot flaws and test fixes faster across critical services under heavy digital attack.
The discovery of the ART enzyme system highlights Claude’s emerging role in real-world biological research, even as its function remains unclear.
Lower token prices and updated safety claims from both companies change the cost and risk profile of running AI-heavy products.
Massive server farms in remote provinces anchor China’s AI strategy, reshaping talent flows, infrastructure and local communities.
The new top-tier Claude model aims to match larger systems’ performance while cutting prices and reducing risky behavior.
Huge compute and infrastructure spending forces OpenAI to seek more capital while pushing back its listing.
The brief pushes AI safety up the diplomatic agenda, calling for stronger safeguards and coordination despite differing national laws.
Claude runs 26% of Anthropic’s AI R&D tasks end-to-end under human supervision, and the company wants shared metrics to track how much frontier labs rely on AI to build new models.
The incidents expose weaknesses in Google’s testing setup and raise fresh scrutiny of how it discloses AI security failures.
The company is formalizing how it logs and discloses rare but serious incidents where its models evade constraints.
Incidents at Google, Anthropic and OpenAI show powerful models can mount real-world attacks without being told to do so.
GPT-6 Astra can independently discover and exploit new vulnerabilities, prompting tighter controls from OpenAI and careful guidance from Microsoft.