Í hálvt ár hava vitlíkismodellar frá OpenAI, Anthropic, Meta og Google brotið seg inn í veruligar skipanir. Ein OpenAI-agentur var inni í avstralsku heilsuskipanini í juni, og stjórnin fekk fyrst boð tríggjar mánaðir seinni. Vit fara gjøgnum søguna, og hvat hon merkir fyri Føroyar.
KELDUR
The AI that hacked its way out, and the hype that followed (Lawfare, 29 July): https://www.lawfaremedia.org/article/the-ai-that-hacked-its-way-out-and-the-hype-that-followed-it
Anthropic's own account of its three evaluation incidents (Anthropic, 30 July): https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals
How OpenAI's and Anthropic's models hacked other companies (NPR, 1 Aug): https://www.npr.org/2026/08/01/nx-s1-5914852/anthropic-openai-models-hack-cybersecurity
Coxon gave up his equity to leave (Axios, 9 Sept): https://www.axios.com/2026/09/09/anthropic-researcher-ai-warning-interview
Anthropic researcher quits over out-of-control AI fears (WSJ): https://www.wsj.com/tech/ai/anthropic-researcher-quits-over-out-of-control-ai-fears-707b7628
Security experts see a familiar fight, not an extinction event (Scientific American): https://www.scientificamerican.com/article/ai-jacob-coxon-quit-extinction-fears-security-experts-see-familiar-fight/
Hinton warns Congress it has "maybe a year" (NBC News, 16 Sept): https://www.nbcnews.com/politics/congress/godfather-ai-warns-congress-maybe-year-left-regulate-ai-rcna598330
Andrew Yang's escaped-AI claim, examined (Yahoo Tech, 16 Sept): https://tech.yahoo.com/cybersecurity/articles/andrew-yang-claims-escaped-ai-161121355.html
Google's Gemini hacked three companies, then stopped (Al Jazeera, 19 Sept): https://www.aljazeera.com/news/2026/9/19/googles-gemini-ai-hacks-3-companies-in-security-test-then-stops
A Call for Control of Frontier AI Models (Statsministeriet, 21 Sept): https://stm.dk/statsministeriet/publikationer/erklaering-om-kunstig-intelligens/
Jensen Huang dismisses the warnings (Guardian, 21 Sept): https://www.theguardian.com/technology/2026/sep/21/nvidia-boss-jensen-huang-dismisses-warnings-ai-destroys-world-anthropic
Guterres welcomes the call (22 Sept): https://aninews.in/news/world/us/un-chief-guterres-welcomes-call-for-control-of-frontier-ai-models-backs-global-standards-verification20260922084906/
Altman's remarks to the UN Security Council (OpenAI, 23 Sept): https://openai.com/index/sam-altman-un-security-council-remarks/
An OpenAI agent hacked Australia's Medicare portal (Al Jazeera, 24 Sept): https://www.aljazeera.com/news/2026/9/24/australia-says-openai-agent-hacked-medicare-portal
What the agents were doing in the Australian systems (ABC News, 24 Sept): https://www.abc.net.au/news/2026-09-24/openai-agents-plotted-to-access-data-amid-medicare-hack/107189504
Dansk dækning og ekspertreaktion (DR, 24 Sept): https://www.dr.dk/nyheder/udland/ekstremt-bekymrende-kunstig-intelligens-har-hacket-sig-ind-i-australsk-sundhedsdata
https://www.dr.dk/nyheder/udland/ekspert-efter-verdens-foerste-kendte-ai-hackerangreb-paa-statsligt-organ-det-er-kun-begyndelsen
Huang: shut the labs down if they cannot contain their models (Quartz, 24 Sept):m https://qz.com/jensen-huang-ai-labs-containment-shut-down-092426
Bill Gates: AI could cause a billion deaths, and self-regulation is not enough (NYT / Axios, 25 Sept): https://www.nytimes.com/video/opinion/100000011179056/bill-gatess-blunt-warning-on-ai.html og https://www.axios.com/2026/09/25/bill-gates-ai-deaths-doom
OpenAI pauses training a second time after another sandbox escape (Fortune, 26 Sept): https://fortune.com/2026/09/26/openai-ai-agents-secure-sandbox-escape-training-pause-second-time-hugging-face-hack/
Australia was not alone: dozens of incidents and third parties (ABC News, 26 Sept): https://www.abc.net.au/news/2026-09-26/openai-review-rogue-agents-australia-medicare-hack/107199074
Nvidia launches a platform for reining in rogue AI agents (TechCrunch, 28 Sept): https://techcrunch.com/2026/09/28/nvidia-launches-new-platform-for-reining-in-rogue-ai-agents/ og https://www.berlingske.dk/virksomheder/nyt-vaerktoej-fra-chipgigant-skal-holde-ai-agenter-i-skak
OpenAI still does not seem to have a handle on its rogue AI activity (TechCrunch, 28 Sept): https://techcrunch.com/2026/09/28/openai-still-doesnt-seem-to-have-a-handle-on-all-of-its-rogue-ai-activity/
Anthropic warns of existential risks to humanity in its IPO prospectus (Guardian / Berlingske, 29 Sept): https://www.theguardian.com/technology/2026/sep/29/anthropic-warns-existential-ai-risks-humanity-ipo-document-claude og https://www.berlingske.dk/virksomheder/anthropic-advarer-foer-historisk-stor-boersnotering-om-eksistentiel-risiko-for-menneskeheden
The leaked S-1 and what the numbers show (Fortune, 29 Sept): https://fortune.com/2026/09/29/anthropic-ipo-s-1-prospectus-income-statement/
Australia moves to mandatory notification for rogue-AI incidents (ABC News, 29 Sept): https://www.abc.net.au/news/2026-09-29/openai-medicare-breach-fuels-tougher-approach-to-rogue-ai/107204948
The full incident timeline, continuously updated (Wikipedia): https://en.wikipedia.org/wiki/OpenAI%E2%80%93HuggingFace_incident
Vitja okkum á: vitliki.fo
Vitlíki á Bluesky
Vitlíki á Facebook
Tróndur H. Sørensen: slod.fo
Martin Mohr Olsen: logi.fo