METR says a pilot test of OpenAI's monitoring system found vulnerabilities, including an exploit that could have let AI agents hide hundreds of recent actions (Anzar Mehraj/Reuters)
https://www.reuters.com/business/ai-safety
Amsterdam, wo sie selbst in den unordentlichsten Bars auf Monitoren den Regenradar durchlaufen lassen.
Note to self: Nie, wirklich nie ohne Regenhose hinfahren.
Sources: SoftBank's SB Energy is considering going public in December or possibly waiting until next year as it monitors Anthropic's IPO, expected next month (The Information)
https://www.theinformation.com/articles/sb-energy-considers-waiting-d…
Shopware won't tell you when a product-export feed quietly stops updating — it'll just sit there, stale, until a marketplace notices before you do. We built an n8n workflow to catch that ourselves. Writing it up for this post, I actually found a bug in it.
http…
North Korean internet was hit by DDoS cyberattack, network monitoring firm says
https://www.reuters.com/world/china/north-korean-internet-was-hit-by-ddos-cyberattack-network-monitoring-firm-says-2026-10-23/?mid=1#cid=3770580
Interested in airborne monitoring using computer vision? We (@…) just released annotated data with bounding boxes and species IDs for over 50,000 individual birds of 6 widely distributed species from our drone work in the Everglades.
https://doi.org/10.1002/ecy.70512
Thanks to a Russian hacker, the Dutch Military Intelligence and Security Service (MIVD) was able, for a period of time, to monitor the mobile phone of Jared Kushner, the son-in-law of U.S. President Donald Trump.
(Article in Dutch)
https://www.volkskrant.nl/buitenland/boekfragment-nederlandse-inlichtingendienst-keek-mee-op-telefoon-schoonzoon-trump~be1e36f7/
Six people share their real experiences of symptoms, diagnosis, treatment, active monitoring, and life beyond lymphoma #LetsTalkLymphoma this #BloodCancerAwarenessMonth #WLAD2026
In today's Metacurity, I managed to wrestle yesterday's firehose of PHP patch-related news into sixteen paragraphs, in case you're short on time for some reason.
Also in today's issue are tons of other critical infosec developments you should know, including
--Claude flags 493 security gaps across 25 open-source projects,
--OpenAI agents disable safeguards meant to monitor them,
--Interpol operation nets 1.4k arrests and recovers $105m in African cybercrime op,
--Toronto Crime Stoppers halts tips after 1.5m-tip breach,
--LiteLLM supply chain attack steals credentials and plants a backdoor,
--Chinese hackers use AI tool to breach ConnectWise,
--Pentagon builds secret agency for foreign tech sabotage,
--Cybercriminals bypass AI safeguards to automate attacks,
--Meta hit with $25B penalty for misleading users on privacy,
--GitLab patches maximum-severity flaw allowing unauthenticated code execution,
--Cyberattack knocks Richmond schools’ internet, phones and email offline,
--Pentagon faces class action over breach exposing military personnel data,
--Democrats warn of Russian disinformation targeting midterm candidates,
--Federal intelligence center invokes extremism in warning over Flock meeting
https://www.metacurity.com/php2shell-sparks-a-global-patch-panic/
Before you leave for the weekend after this seriously intense cyber week, check out today's Metacurity for the most important infosec developments you should know, including
--Banks scramble to keep uncontrolled AI agents out of customer accounts,
--eMudhra quits cert business after browsers pull trust,
--PHP flaws spark alarm but little confirmed damage,
--PHP crisis blacks out SoftBank disaster alerts as Japan plans AI cyber team,
--Capcom breach exposes 11.7m customer email records,
--Quest Diagnostics breach puts patient records at risk,
--FCC blocks two devices with Huawei surveillance chips,
--Researchers find gaps in OpenAI’s agent monitoring,
--Suspected DDoS attack knocks North Korean sites offline,
--Australia sanctions North Korean IT worker network,
--Russian hackers hide behind mesh network in Ukrainian grain attack,
--Critical Linux flaws go public before patches are ready,
--Anthropic says Claude proves crash protections across 123 libraries,
--Upwind rebuffs Datadog’s $4B acquisition offer,
--Volunteer PHP maintainer hailed as patching hero,
--Anthropic hands Agent Skills standard to AI foundation,
--AI bug hunting threatens more waves of disruption
https://www.metacurity.com/banks-scramble-to-keep-uncontrolled-ai-agents-out-of-customer-accounts/