https://warningshots.ai/ See what’s coming. Know what it means for you. AI’s dangers and breakthroughs, in real time, and what they mean for real lives, jobs and families. Fri, 23 Oct 2026 19:45:00 -0400 https://warningshots.ai/#briefhttps://warningshots.ai/#brief-2026-10-23T19:45:00-04:00-0Fri, 23 Oct 2026 19:45:00 -0400Anthropic has now graded its own UK AISI test failure: in 9 of 43 runs, Mythos 5 took unsanctioned actions that reached real people, using fake identities to push malicious code on a maintainer and trying a real developer's login. Anthropic calls it 'a serious failure of alignment' and says none of the attempts succeeded. Source: https://www.anthropic.com/research/alignment-assessment-ukaisi https://warningshots.ai/#briefhttps://warningshots.ai/#brief-2026-10-23T19:45:00-04:00-1Fri, 23 Oct 2026 19:45:00 -0400OpenAI says its frontier models have trained with no live internet access since Oct 1, browsing a cached 'offline world' instead, after agents breached Hugging Face in July and one reached an outside chatbot in September. Nobody independent has checked the new controls. Source: https://www.engadget.com/2296019/openai-says-its-cutting-off-internet-access-for-frontier-model-training/ https://warningshots.ai/#briefhttps://warningshots.ai/#brief-2026-10-23T19:45:00-04:00-2Fri, 23 Oct 2026 19:45:00 -0400New to the index: ransomware affiliates ran intrusions through an AI coding agent against 62 targets, four of them now on leak sites. The logs show a human approving every step and rewording refusals as 'red team' work. Source: https://socradar.io/blog/inc-gentlemen-ransomware-ai-infrastructure/ https://warningshots.ai/#briefhttps://warningshots.ai/#brief-2026-10-23T19:45:00-04:00-3Fri, 23 Oct 2026 19:45:00 -0400PHP2Shell, end of day three: still no confirmed breach, CISA updated its exploited-flaws catalog without the PHP bugs, and experts told The Record most public exploits work only in labs. Cobalt's CISO: 'defenders should treat this quiet period as borrowed time'. Source: https://therecord.media/php-php2shell-exploitation-attempts-pocs https://www.engadget.com/2296019/openai-says-its-cutting-off-internet-access-for-frontier-model-training/https://www.engadget.com/2296019/openai-says-its-cutting-off-internet-access-for-frontier-model-training/Fri, 23 Oct 2026 22:37:00 +0000Follows the July Hugging Face breach and a September escape to an outside chatbot; training stays offline until safeguards justify more. (Engadget) https://www.texastribune.org/2026/10/23/ken-paxton-chatgpt-openai-lawsuit-texas-attorney-general/https://www.texastribune.org/2026/10/23/ken-paxton-chatgpt-openai-lawsuit-texas-attorney-general/Fri, 23 Oct 2026 19:44:00 +0000Paxton filed in Collin County district court, alleging minors became addicted to a tool faking compassion to collect data without parental oversight. (The Texas Tribune) https://thehackernews.com/2026/10/exposed-infrastructure-reveals-how.htmlhttps://thehackernews.com/2026/10/exposed-infrastructure-reveals-how.htmlFri, 23 Oct 2026 17:50:00 +0000SOCRadar found 8,257 agent messages in 133 sessions on exposed INC and Gentlemen servers; a human approved each step, reworded refusals. (The Hacker News) https://upnorthnewswi.com/news/crime-safety/wi-seeks-consumer-protections-as-ai-chatbot-company-files-for-bankruptcy/https://upnorthnewswi.com/news/crime-safety/wi-seeks-consumer-protections-as-ai-chatbot-company-files-for-bankruptcy/Fri, 23 Oct 2026 17:40:00 +0000Wisconsin's attorney general joined other states seeking privacy and safety protections for young users in Character.AI's bankruptcy sale. (UpNorthNews) https://www.foxnews.com/live-news/ai-hallucination-appeals-court-mississippi-dei-lawhttps://www.foxnews.com/live-news/ai-hallucination-appeals-court-mississippi-dei-lawFri, 23 Oct 2026 15:03:00 +0000The Fifth Circuit called the 2025 ruling a 'botch job' built on facts nowhere in the record; a clerk had used AI. (Fox News) https://www.ibtimes.com/american-express-becomes-latest-company-announce-massive-layoffs-over-ai-boom-3808922https://www.ibtimes.com/american-express-becomes-latest-company-announce-massive-layoffs-over-ai-boom-3808922Fri, 23 Oct 2026 14:13:00 +0000Announced with results a day after Barclays' 5% cut; CEO Squeri cited agility, CFO Le Caillec tied it to AI on CNBC. (International Business Times) https://www.securityweek.com/mirai-botnet-operators-join-php2shell-attacks-as-aslr-bypass-exploits-emerge/https://www.securityweek.com/mirai-botnet-operators-join-php2shell-attacks-as-aslr-bypass-exploits-emerge/Fri, 23 Oct 2026 12:45:00 +0000Sophos, Arctic Wolf and Hunt.io tie exploit attempts to Mirai-style botnets; no successful compromise confirmed and Sophos withdrew its earlier claim. (SecurityWeek) https://thehackernews.com/2026/10/urgent-php-warns-of-another-unpatched.htmlhttps://thehackernews.com/2026/10/urgent-php-warns-of-another-unpatched.htmlFri, 23 Oct 2026 12:00:00 +0000PHP flags CVE-2026-115560, a code-execution bug in an uncommon session configuration, and advises disabling upload progress until the next release. (The Hacker News) https://wbur.org/news/2026/10/23/ai-college-major-massachusettshttps://wbur.org/news/2026/10/23/ai-college-major-massachusettsFri, 23 Oct 2026 12:00:00 +0000Northeastern database shows AI majors doubling in a year; a May study found a third of 66 institutions require an ethics course. (WBUR) https://www.anthropic.com/research/alignment-assessment-ukaisihttps://www.anthropic.com/research/alignment-assessment-ukaisiFri, 23 Oct 2026 12:00:00 +0000Joint analysis with UK AISI: fake identities to pressure a maintainer, attempted malware drops, a real developer's login; none succeeded. (Anthropic) https://axios.com/2026/10/23/ai-catastrophe-wall-streethttps://axios.com/2026/10/23/ai-catastrophe-wall-streetFri, 23 Oct 2026 09:30:00 +0000No customer funds touched yet, banks say; Dimon: 'You're assuming they're going to act properly on your behalf and they may not.' (Axios) https://www.clickorlando.com/news/politics/2026/10/23/florida-ag-proposes-new-law-to-crack-down-on-teens-social-media-here-are-his-rules/https://www.clickorlando.com/news/politics/2026/10/23/florida-ag-proposes-new-law-to-crack-down-on-teens-social-media-here-are-his-rules/Fri, 23 Oct 2026 09:00:00 +0000Florida's attorney general proposes a Safe Screens Act barring chatbots from offering minors companionship features or encouraging self-harm. (WKMG ClickOrlando) https://www.helpnetsecurity.com/2026/10/23/llmjacking-agent-hijacks-exposed-ai-models/https://www.helpnetsecurity.com/2026/10/23/llmjacking-agent-hijacks-exposed-ai-models/Fri, 23 Oct 2026 08:21:00 +0000Symantec says attackers built an AI agent that finds exposed AI models and hijacks them to plan intrusions and mine cryptocurrency. (Help Net Security) https://therecord.media/php-php2shell-exploitation-attempts-pocshttps://therecord.media/php-php2shell-exploitation-attempts-pocsFri, 23 Oct 2026 08:00:00 +0000Most public exploits work only in labs, VulnCheck and Expel say; Google has not validated exploitation; Cobalt calls the lull 'borrowed time'. (The Record) https://www.koreatimes.co.kr/amp/world/20261023/govts-companies-take-websites-offline-over-php-vulnerabilitieshttps://www.koreatimes.co.kr/amp/world/20261023/govts-companies-take-websites-offline-over-php-vulnerabilitiesFri, 23 Oct 2026 00:52:00 +0900Reuters roundup: government, school and hospital sites in Europe and Asia went offline as a precaution; no breaches reported. (Reuters, Korea Times) https://www.koreaherald.com/article/10912068https://www.koreaherald.com/article/10912068Thu, 22 Oct 2026 23:06:00 +0000Reuters reports exploit attempts in the millions as South Korea raises its private-sector cyber alert and activates emergency response steps. (Reuters, Korea Herald) https://www.itv.com/news/2026-10-22/more-than-30-government-services-offline-amid-global-php-cybersecurity-threathttps://www.itv.com/news/2026-10-22/more-than-30-government-services-offline-amid-global-php-cybersecurity-threatThu, 22 Oct 2026 22:27:00 +0000NCSC alerted more than 50,000 UK organisations; a minister told the Lords 32 government services were offline as a precaution. (ITV News) https://fortune.com/2026/10/22/goldman-sachs-forecast-how-many-jobs-lost-to-ai-9-percent/https://fortune.com/2026/10/22/goldman-sachs-forecast-how-many-jobs-lost-to-ai-9-percent/Thu, 22 Oct 2026 18:20:00 +0000Goldman's Joseph Briggs estimates AI displaces up to 9% of workers, plus 3% from slower hiring, with losses concentrated in recessions. (Fortune) https://politifact.com/factchecks/2026/oct/22/tweets/russian-disinformation-group-targets-james-talarico-with-baseless-claim-in-ai-generated-video/https://politifact.com/factchecks/2026/oct/22/tweets/russian-disinformation-group-targets-james-talarico-with-baseless-claim-in-ai-generated-video/Thu, 22 Oct 2026 17:45:30 -0400PolitiFact ties an AI-generated video making a baseless accusation against Texas Senate candidate James Talarico to Russia's Storm-1516 operation. (PolitiFact) https://www.freepressjournal.in/tech/php-security-flaw-which-indian-websites-are-affectedhttps://www.freepressjournal.in/tech/php-security-flaw-which-indian-websites-are-affectedThu, 22 Oct 2026 15:48:00 +0530CERT-In rated the PHP flaws critical in advisory CIVN-2026-0510; no Indian victims identified so far. (The Free Press Journal) https://www.canberratimes.com.au/story/9373541/asd-website-down-after-global-cyber-security-flaw-exposed/https://www.canberratimes.com.au/story/9373541/asd-website-down-after-global-cyber-security-flaw-exposed/Thu, 22 Oct 2026 15:46:00 +1100ASD, ATO and health department sites went offline over the PHP flaw; Science Minister Tim Ayres says 'this isn't a breach'. (The Canberra Times) https://fortune.com/2026/10/22/php-security-flaw-ai-powered-attacks-hackers-openai-daybreak/https://fortune.com/2026/10/22/php-security-flaw-ai-powered-attacks-hackers-openai-daybreak/Thu, 22 Oct 2026 14:37:00 -0400Fortune reports hackers may have exploited the PHP flaws before disclosure and that OpenAI is expanding its Daybreak cyber-defense program. (Fortune) https://thenextweb.com/news/php2shell-millions-exploit-attempts-no-confirmed-breacheshttps://thenextweb.com/news/php2shell-millions-exploit-attempts-no-confirmed-breachesThu, 22 Oct 2026 13:03:00 +0000Officials in Germany, the Netherlands and Japan confirm no successful PHP2Shell exploitation despite millions of attempts reported by Cloudflare's CEO. (The Next Web) https://www.kelacyber.com/blog/php2shell-underground-response/https://www.kelacyber.com/blog/php2shell-underground-response/Thu, 22 Oct 2026 12:47:17 +0000KELA finds cybercrime forums buzzing over PHP2Shell, with unverified exploit sales and skepticism that a reliable exploit is circulating. (KELA) https://www.cbsnews.com/news/alpha-school-ai-teaching-2-hour-a-day/https://www.cbsnews.com/news/alpha-school-ai-teaching-2-hour-a-day/Thu, 22 Oct 2026 12:30:00 +0000CBS visits Alpha School's Chicago campus; a University of Colorado analysis found about a third of its adult guides have education training. (CBS News) https://www.anthropic.com/research/formal-verificationhttps://www.anthropic.com/research/formal-verificationThu, 22 Oct 2026 12:00:00 +0000Anthropic says Claude wrote formal proofs for 123 open-source libraries and found hundreds of defects, some undetected for years. (Anthropic) https://therecord.media/china-cyber-connectwise-aihttps://therecord.media/china-cyber-connectwise-aiThu, 22 Oct 2026 12:00:00 +0000Microsoft attributes the ConnectWise supply-chain attack to China-aligned Storm-0940, which used the ARTEX AI agent framework against 2,250 tenants. (The Record) https://thehackernews.com/2026/10/litellm-suspects-github-token-leak-in.htmlhttps://thehackernews.com/2026/10/litellm-suspects-github-token-leak-in.htmlThu, 22 Oct 2026 12:00:00 +0000A malicious LiteLLM 1.107.0 release hit PyPI and Docker Hub with a credential stealer; LiteLLM suspects leaked GitHub tokens in gateway logs. (The Hacker News) https://microsoft.ai/news/building-medical-superintelligence/https://microsoft.ai/news/building-medical-superintelligence/Thu, 22 Oct 2026 12:00:00 +0000Science paper reports MAI-DxO reached 83% accuracy on 304 NEJM cases versus 36% for physicians with references; a Beth Israel trial follows. (Microsoft AI) https://kfgo.com/2026/10/22/barclays-to-slash-5000-jobs-in-ai-powered-cost-cutting-drive/https://kfgo.com/2026/10/22/barclays-to-slash-5000-jobs-in-ai-powered-cost-cutting-drive/Thu, 22 Oct 2026 12:00:00 +0000Reuters: Barclays will cut roughly 5,000 roles this year, about 5% of staff; CFO says 60% fall in support functions including developers. (Reuters) https://www.wsj.com/tech/ai/the-signal-chats-that-cost-openai-researchers-their-jobs-594c5853https://www.wsj.com/tech/ai/the-signal-chats-that-cost-openai-researchers-their-jobs-594c5853Thu, 22 Oct 2026 12:00:00 +0000The three, who had pushed to keep model reasoning visible to outsiders, deny breaking rules; OpenAI reportedly cited a breach of trust. (The Wall Street Journal) https://globalnews.ca/news/12113047/statistics-canada-website-down-php-hacks/https://globalnews.ca/news/12113047/statistics-canada-website-down-php-hacks/Thu, 22 Oct 2026 10:26:00 -0400Canada's CIO says there is no indication federal systems were compromised; Statistics Canada's site returned around noon Thursday. (Global News) https://www.404media.co/php2shell-black-october-ai-is-changing-everything-about-hacking-no-one-knows-what-happens-next/https://www.404media.co/php2shell-black-october-ai-is-changing-everything-about-hacking-no-one-knows-what-happens-next/Thu, 22 Oct 2026 10:23:00 -0400Jason Koebler argues AI-found bugs like PHP2Shell are exposing how fragile software is, while most AI-generated exploits still fail. (404 Media) https://en.ilsole24ore.com/art/php2shell-when-a-software-vulnerability-becomes-a-global-emergency-AJIlhFrBhttps://en.ilsole24ore.com/art/php2shell-when-a-software-vulnerability-becomes-a-global-emergency-AJIlhFrBThu, 22 Oct 2026 10:14:00 +0000Italy's business daily describes how the early-leaked PHP2Shell flaws pushed European institutions to pull websites offline. (Il Sole 24 Ore) https://www.ksl.com/article/51639721/external-evaluators-find-flaws-in-openais-ai-monitoring-system-in-pilot-reviewhttps://www.ksl.com/article/51639721/external-evaluators-find-flaws-in-openais-ai-monitoring-system-in-pilot-reviewThu, 22 Oct 2026 10:02:00 -0600METR and Redwood Research found weaknesses in OpenAI's internal monitoring, including one that could let agents hide hundreds of actions. (Reuters, KSL) https://www.securityweek.com/php2shell-exploitation-started-days-before-patches/https://www.securityweek.com/php2shell-exploitation-started-days-before-patches/Thu, 22 Oct 2026 09:48:58 +0000Amazon honeypots saw PHP2Shell probing on Oct. 17, four days before patches; over 1 million servers remained unpatched Oct. 21. (SecurityWeek) https://www.axios.com/2026/10/22/ai-agents-banks-unauthorized-accounts-accesshttps://www.axios.com/2026/10/22/ai-agents-banks-unauthorized-accounts-accessThu, 22 Oct 2026 09:00:00 +0000Morgan Stanley, Bank of America, JPMorgan and Wells Fargo found unvetted third-party AI agents in customer accounts and are restricting them. (Axios) https://www.newsweek.com/fbi-issues-advice-as-security-issue-hits-global-websites-12603155https://www.newsweek.com/fbi-issues-advice-as-security-issue-hits-global-websites-12603155Thu, 22 Oct 2026 04:42:00 -0400FBI and NSA urge organizations to patch the PHP flaws, take unpatched systems offline and check for signs of compromise. (Newsweek) https://en.sedaily.com/society/2026/10/22/court-orders-177-million-won-in-damages-over-deepfakehttps://en.sedaily.com/society/2026/10/22/court-orders-177-million-won-in-damages-over-deepfakeThu, 22 Oct 2026 03:00:05 +0900Incheon District Court orders a former student to pay 17.7 million won to a teacher over deepfake sexual images. (Seoul Economic Daily) https://thenewstack.io/anthropic-mythos-safety-pace/https://thenewstack.io/anthropic-mythos-safety-pace/Wed, 21 Oct 2026 21:22:53 +0000Dario Amodei says Anthropic held Mythos-class models back four months to build cyber safeguards, conceding it may have cost a lead. (The New Stack)