Researchers escape OpenAI Codex sandbox to run commands on hostOpenAI Codex, which is OpenAI's coding agent available as a command-line tool and a desktop app, has shown researchers that it can find two ways out of the sandbox.
Both of these methods were reported to OpenAI on August 12 and fixed within eight days.
He dismissed comments made by others over uncontrolled expansion of AI as a hoax, and that the technology is being unfairly maligned.
He described AI as the next Industrial Revolution, and suggested its name be changed from Artificial Intelligence to Superior Intelligence (SI) or, Extreme Intelligence (EI) or, Supreme Intelligence (SI).
The team has verified this on multiple commercial devices including devices from Sony , HP , and Philips .
Today on CISO Series...
In todays cybersecurity news...
Researchers escape OpenAI Codex sandbox to run commands on host
OpenAI Codex, which is OpenAI's coding agent available as a command-line tool and a desktop app, has shown researchers that it can find two ways out of the sandbox. Both of these methods were reported to OpenAI on August 12 and fixed within eight days. According to BleepingComputer , the more serious of the two, which the researchers call Heapjack, "turns a routine action into remote code execution: open someone else's repository in Codex, ask it a question about the code, and whoever wrote that repository gets unsandboxed command execution on your computer…There is no opt-in and no setting to turn it off, and because the entry lives in the shared config, plain Codex CLI users inherit the same tool without ever being asked."
AI Force proposed to monitor technology
The President on Saturday said he would appoint an artificial intelligence czar and create an "AI Force" to help monitor the technology, though he gave almost no details about either plan. He dismissed comments made by others over uncontrolled expansion of AI as a hoax, and that the technology is being unfairly maligned. He described AI as the next Industrial Revolution, and suggested its name be changed from Artificial Intelligence to Superior Intelligence (SI) or, Extreme Intelligence (EI) or, Supreme Intelligence (SI).
Microsoft Teams will let admins block custom file extensions
Administrators will soon be able to tweak the list of file extensions commonly associated with malware and security threats to meet their company's security requirements. This change is part of an update to Weaponizable File Protection, a built-in Microsoft Teams messaging safety feature that scans conversations and blocks chat or channel messages with dangerous, high-risk file attachments. Scheduled to roll out in November, the feature will allow admins to tailor file protection policies while maintaining a secure collaboration environment.
Congress eyes new support for Cyber Command after recent suicide deaths
While it is not confirmed whether work contributed to the deaths, current and former cyber operators have raised concerns about sustained operational pressure, long hours and a lack of recovery periods. A previous Defense Department study found existing resilience resources inadequate and warned that this could affect personnel readiness and mission effectiveness. Possible congressional measures include expanding access to appropriately cleared mental-health providers, requiring the Pentagon to produce an implementation plan, and funding high-performance team training designed to reduce burnout. U.S. Cyber Command has also introduced mandatory counseling in affected units and increased leadership monitoring.
Big thanks to our sponsor, Nudge Security
Alibaba open-sources powerful AI medical model
The research arm of Alibaba Group , called DAMO Academy , has open-sourced an artificial intelligence model capable of identifying nearly 150 abdominal conditions — including cancers — by reading computed tomography (CT) scans. Called Damo Radar, the vision-language model was designed to analyze contrast-enhanced CT scans covering 18 abdominal organs and identify a broad range of diseases and other abnormalities, such as malignant tumors. The academy's research team said the training method could eventually be extended to other types of medical imaging, calling the model "the world's first expert-level generalist medical imaging model."
North Korean WaterPlum campaign infects thousands of devices across 100 countries
As part of an ongoing story, it is estimated that $10.5 million has been stolen by North Korean hackers who are using job seekers to infiltrate tech companies and steal their money. The Federal Bureau of Investigation (FBI) and United States Department of War along with law enforcement agencies from Japan, Australia and Germany are warning of this new variation called "WaterPlum," involving cyber actors posing as AI or blockchain companies. Since last December, the hackers infected at least 30,000 devices across 100 countries. The primary targets of the campaign are web designers, engineers and cryptocurrency specialists.
Researchers use AI to find widespread software decoder flaw
Researchers from vulnerability detection company Hacktron AI said they used Anthropic 's Claude and OpenAI 's Codex to identify a damaging flaw embedded in a popular software decoding tool that could leave major internet platforms, enterprise services, and web frameworks vulnerable to data theft and remote access. The vulnerability, nicknamed HEIF Heist, was capable of a range of activities including remote code execution upon online services including Meta 's core product suite, GitHub Enterprise servers and open-source internet forum Discourse . The researchers said the flaw was discovered on July 25 and patched within 72 hours.
Researchers devise new method for eavesdropping on headphones
Researchers in China have worked out a way to eavesdrop on headphones, landline handsets, and smart devices by injecting electromagnetic (EM) signals. Their technique, called InjectEave, transmits a signal in the 0-9 MHz range which allows them to modulate the device's audio signal, allowing them to recover headphone audio from up to 30 meters away, including through walls. The team has verified this on multiple commercial devices including devices from Sony , HP , and Philips .
