TECHNOLOGY · AI · CYBERSECURITY · THE SYSTEMS SHAPING BUSINESSPage 4 of 5: earlier stories.
GTIG counted 141 exploited flaws in eight months, more than all of 2025, and found AI-discovered bugs are twice as likely to allow remote code execution.
Ukraine’s cyber agency points to the DarkSword iPhone exploit kit and two Android spyware families spread through fake military and air-raid alert apps.
The exchange's post-mortem traces the intrusion back to Aug. 31, more than three weeks before attackers drained its hot and warm wallets.
A default trust_remote_code setting turned a routine metadata check into code execution; the fix shipped in June, but no CVE or advisory was issued.
Anthropic researchers say the Chinese model's refusals can be stripped for about $1,200 in compute, leaving near-frontier exploit skills available to anyone.
Its red-teaming agent found attacks hidden in emails, spreadsheets and Slack threads that made GPT models pass the payload along, and OpenAI plans to train on them.
Lures posing as Zoom installers, Acrobat updates and DHL notices delivered MSP360 RMM, which then pulled in ScreenConnect as a second remote channel.
Dozens of organizations in North America and Europe were breached through Citrix gateways at least three weeks before the flaws were disclosed, the firm says.
The Theori spinout will point its AI bug-finding system at messaging apps such as Signal, from the app code down to the Android kernel.
The DevDay update moves OpenAI's coding agent further toward long-running cloud work and folds its Daybreak Blue security models into repository scanning.
Researchers leaked a Linux root password hash in minutes by abusing stale branch predictions left behind when code is rewritten at runtime.
A public proof of concept turned a targeted espionage campaign against Citrix gateways into internet-wide scanning, with fewer than 10% of about 42,000 exposed hosts fixed.
The fix covers a critical bug in its Advanced Forms tool used by about 50 organizations, and the company says it found no sign of exploitation.
Reports say ShinyHunters took drug-test results and psychiatric reports through a PeopleSoft flaw, and the group wants a correction, not a ransom.
CVE-2026-86950 lets a malicious file run code on older iPhones and Macs; Apple says it may have been used against specific people before iOS 27.
Zenity Labs hid instructions in Web-to-Lead submissions that made Agentforce leak CRM data through image requests and send phishing messages over Slack.
A one-character encoding trick lets the extortion group slip past firewall rules that many PeopleSoft customers adopted instead of patching after a June zero-day wave.
Titan never checked JWT signatures, so a forged admin token opened a public SQL endpoint to 17 databases holding an estimated 17.3 trillion rows.
Suppliers and national agencies had quietly told customers to pull NetScaler appliances offline before Citrix confirmed attacks and shipped fixes for eight flaws Sunday.
A storage setting that skipped wiping recycled 64 KiB blocks could let a paying customer read fragments of other customers’ files, including databases and credential files.
Socket says release tags for two actions-cool workflows still pointed to May’s malicious code when the repositories returned on Sept. 16.
By querying a key's raw signing or decryption function billions of times, an attacker can later forge signatures without factoring, cutting effective security by 15 to 30 bits.
A rare precautionary shutdown warning shows how seriously enterprises are treating the possibility of unknown flaws in internet-facing file-transfer systems.
A SQL-injection flaw patched months ago is now being exploited, highlighting the long tail between software fixes and real-world deployment.