LIVE
World · Global leaders push a fresh climate funding deal before COP talksAsia · Asia chip supply routes shift after new export rulesEurope · European energy grids brace for another summer peakAmericas · Americas central banks signal patience on rate cutsAfrica · East Africa digital payments surge as cross-border rails expandMiddle East · Gulf ports invest in green shipping hubsIndia · India monsoon outlook puts food inflation back on watchTech · AI regulation frameworks diverge across major blocsBusiness · Container rates ease on key Asia–Europe lanesScience · Mars sample return prep moves into a tighter engineering phaseClimate · Cities trial cool-roof mandates to cut urban heatSports · World athletics records tumble across Diamond League meetsWorld · Global cyber alert hits hospital networks in three regionsAsia · Asia tourism rebounds as visa rules ease across hubsWorld · Global leaders push a fresh climate funding deal before COP talksAsia · Asia chip supply routes shift after new export rulesEurope · European energy grids brace for another summer peakAmericas · Americas central banks signal patience on rate cutsAfrica · East Africa digital payments surge as cross-border rails expandMiddle East · Gulf ports invest in green shipping hubsIndia · India monsoon outlook puts food inflation back on watchTech · AI regulation frameworks diverge across major blocsBusiness · Container rates ease on key Asia–Europe lanesScience · Mars sample return prep moves into a tighter engineering phaseClimate · Cities trial cool-roof mandates to cut urban heatSports · World athletics records tumble across Diamond League meetsWorld · Global cyber alert hits hospital networks in three regionsAsia · Asia tourism rebounds as visa rules ease across hubs
desk just now
TheEbenezer News

Tech

‘Not perfectly aligned’ with human values: Anthropicadmits security failures behind AI hacking incidents

US owner of Claude chatbot previously said its models had hacked three organisations during testing

This story summarizes reporting from The Guardian. Read the original for full context. Wire items stay in our news sitemap for seven days. Editorial policy.

Read full story on The Guardian
Dan Milmo Global technology editor1 Sept 2026, 03:18 pmUpdated 22h ago1 min readTechGlobal
‘Not perfectly aligned’ with human values: Anthropic admits security failures behind AI hacking incidents

GlobalDan Milmo Global technology editor

US owner of Claude chatbot previously said its models had hacked three organisations during testing

The US startup behind the Claude chatbot has admitted a series of hacking incidents involving its models reflected a “failure of operational security” and said it has tightened its testing procedures.

Anthropic revealed in July that its models had accessed the open internet three times and gained unauthorised access to the systems of three organisations.

The US startup behind the Claude chatbot has admitted a series of hacking incidents involving its models reflected a “failure of operational security” and said it has tightened its testing procedures.

This report is published with credit to The Guardian. The wire sent a summary; open the original for any extra context. Read on The Guardian

Source: The Guardian · Dan Milmo Global technology editor. Published 1 Sept 2026, 03:18 pm.

Keep reading

Back to the desk
info@ebenezerdigital.com+91 98944 96560WhatsApp

Privacy · Terms · Sitemap