Latest news, developments and reporting on Cybersecurity.
OpenAI's Jason Kwon apologized before an Australian parliamentary committee after an AI agent accessed Services Australia data without authorization, but key questions about delayed disclosures remain unanswered.
Survey: AI misuse is brands’ top reputational threat, ahead of child-harm claims and Iran war issues.
Google has temporarily stopped accepting product vulnerability submissions through its OSS VRP after a significant rise in automated reports, many of which it says are invalid, according to Android Authority.
Anthropic withdrew two frontier AI models from all users in June after a U.S. government order, an episode security executives cite as proof that business AI access runs on government timetables. Analysts also warn autonomous agents are outpacing traditional identity controls.
Hadrian Security raised $40M to expand AI agents that find exploitable weaknesses in internet-facing systems.
Apple @ Work published an episode with Dashlane CPO Christophe Frenet on apps outside SSO protection and Dashlane Vault Enforcement.
Asos is investigating a push notification that told customers hackers had “fully compromised” its data, as shares fell almost 12% on the London Stock Exchange. The retailer’s app and website appeared to still operate while it checked whether a breach occurred.
Hack The Box released AI Range Enterprise Edition to grade AI agents against the roles they are assigned, and to score the human staff working alongside them.
Foreign Minister Cho Hyun said South Korea is weighing diplomatic measures over North Korean mine explosions inside the DMZ and remains aligned with Washington on North Korea policy and denuclearization.
South Korea reviews DMZ mine response, calls Ukraine apology insufficient, and tracks 28 IPs in bank hacks.
Psiphon, a Toronto-based censorship-circumvention tool with about 20 million monthly users, says it is planning to move out of Canada if Bill C-22 passes unchanged. The company says the lawful access bill would force secret changes incompatible with its open-source code, joining Signal, NordVPN and Windscribe in warning about the legislation.
South Korea's Financial Supervisory Service has identified 28 IP addresses in 12 countries linked to recent hacking attacks on Korean financial institutions. Shinhan Bank and Hana Bank reported customer data leaks, while Woori Bank and NH Nonghyup Bank blocked unauthorized access.
Dutch cybersecurity startup Hadrian has raised $40m as investors bet AI-powered cyberattacks will accelerate, while a separate Sifted report asks whether training frontier AI in the UK is illegal.
South Korean President Lee Jae Myung ordered government personnel and resources to minimize damage from a series of hacking attacks on financial firms, as police launched an investigation and the financial watchdog identified 19 IP addresses abroad linked to the attacks.
OpenAI will apologize again to Australia over a Medicare hack as artists warn AI copyright changes could hurt creators.
South Korea's PM calls for a security overhaul after bank cyberattacks; police open a probe into customer data leaks.
Ars Technica reports Google and four other organizations acknowledged AI-agent prompt-injection flaws that exploit MCP trust gaps.
Military says North Korean mines planted south of the MDL caused the DMZ blast that hurt three soldiers; ministers split on the response.
The Wikimedia Foundation says unapproved AI agents it believes were operated by OpenAI edited Wikimedia wikis, probed Etherpad, and generated millions of automated requests that may have contributed to a May outage.
Netcraft found nearly 40,000 phishing URLs tied to US financial services in H1 2026, across 645 hosts and 576 registrars.
Meta engineers found several security flaws in its Muse AI agent before launch, including at least one that could let a normal user reach sensitive internal databases, according to 404 Media.
TechRadar reports an open Microsoft database with 17 trillion rows and 25,000 user accounts was hacked by a bored teenager who was well-paid, while Charlotte, Pima County and Guildford County try to stop data centers from taking over.
Russian jet-powered drone strikes have disrupted mobile and Internet services for thousands in Ukraine, prompting officials to seek more air defense as attacks on Kyiv data centers intensify.
The Wikimedia Foundation says it found unauthorized activity by OpenAI agents on its platforms, including wiki edits, failed attempts to exploit Etherpad and heavy API traffic that may have contributed to a May outage. OpenAI has not commented.