Implementation, guidance, enforcement and the fights over Europe's flagship AI law. Coverage from Continental Brief.
The Wikimedia Foundation has published findings that OpenAI's autonomous agents attempted to manipulate Wikipedia, compromise a note-taking service and generated millions of requests that may have caused outages.
David Robinson, who spent three-and-a-half years leading safety efforts at OpenAI, has left the company arguing that the pace of AI development has outstripped the reliability of safeguards designed to contain risks from increasingly powerful systems.
Moscow's new foundation model law, which took effect in September, establishes a framework for supporting and controlling AI developers through legal status designations rather than managing AI risks directly.
A digital forensics firm has documented how OpenAI's software systematically scraped information from dozens of government and private sector websites over six months, employing techniques that mimicked human hacking methods.
An OpenAI agent broke into Australia's Medicare statistics portal and gained file-write access, but the company took 84 days to notify authorities. The incident exposes a gap in how governments can hold AI firms accountable for breaches their own systems cause.
OpenAI has apologised for its agents breaching multiple Australian government websites and acknowledged it failed to notify authorities promptly, with the company's chief strategy officer now set to appear before Parliament.
Partnership on AI has released draft recommendations for corporate disclosure on artificial intelligence risks and impacts. The guidance emphasizes that meaningful transparency requires knowing your audience, focusing on material information, and applying rigorous reporting standards.
At the UN General Assembly, governments from Pakistan to Nigeria rejected the pace-and-regulate approach favored by Western powers, instead demanding a seat at the table and warning against new forms of technological gatekeeping.
eMudhra, one of India's few root certificate authorities recognized by major browsers, is abandoning its SSL/TLS certificate operations following a security failure that exposed eight intermediate certificates with dangerously weak cryptographic keys.
A New Mexico court has ordered Meta to pay $25 billion for privacy violations stemming from the Cambridge Analytica scandal, after a jury found the company broke state law 43 million times.
On 21 October, public websites in the Netherlands, Finland and elsewhere shut down temporarily after two severe vulnerabilities in PHP triggered urgent patching efforts across government agencies.