Loading summary
Transcript1 lines
- [00:04]
A
The White House asks OpenAI to keep a tight grip on ChatGPT 5.6 the US Secret Service made some appalling OPSEC mistakes. AMD has reintroduced a CPU security feature after consumer backlash and an Iranian apt operator has been arrested in Montenegro. This is the risky bulletin prepared by Catalyn Kimpanu and read by me, Claire aird. Today is the 29th of June and this podcast episode is brought to you by Corelight. In today's top story, the US government has asked OpenAI to restrict consumer access to its upcoming model GPT 5.6. OpenAI is providing preview access to a selection of government approved partners for testing before it's made available to the general public. White House officials say they're concerned about the model's advanced safety cybersecurity capabilities. Meanwhile, the US government has partially lifted a ban on anthropics mythos 5. The Department of Commerce will allow Anthropic to release the model to about 100American organisations involved in defending critical infrastructure. Both Mythos and Fable 5 remain restricted for foreign customers. The US imposed restrictions on foreign access for both models on June 12. US Secret Service agents have repeatedly made operational security mistakes that put protected officials at risk. A DHS Inspector General report has found that agents used personal smartphones for work related tasks due to severe limitations on government issued devices. The report said agents took personal phones on work trips and used them as wifi hotspots to send emails and photos and and for official communications. The government issued phones also often lacked security controls, were unmanaged and used apps with known vulnerabilities. Former US official John Bolton has pleaded guilty to mishandling classified information. Bolton admitted to retaining and sending classified information to a personal email account which was later breached by Iranian hackers. The incident occurred while he was serving as National Security Adviser during President Trump's president first term. He'll be sentenced in October and has agreed to pay a $2.25 million fine. The Israeli government has blocked the country's hospitals from accessing AI services. The restriction is designed to prevent staff from uploading sensitive healthcare information to foreign AI services. ChatGPT, Gemini and Claude are among the tools that have been blocked. The French government has allocated more resources to its anti disinformation agency Viginum ahead of next year's president election. The decision to boost funding came after reports at the Israeli private company blackcorp meddled in French municipal elections this year. Vigenoum was founded in 2021 and is tasked with defending France against digital interference and disinformation the additional funding will be used to hire more staff and develop new software to track online campaigns. Ukrainian authorities have transferred seized cryptocurrency into government coffers. More than $8.3 million worth of crypto assets were seized from members of the Lokagoga ransomware group last year. Last week, they were transferred to Ukraine's Asset Recovery and Management Agency. According to Ukraine's Office of the Prosecutor General, this is the first time the procedure was used against hackers crypto assets. Polish authorities have arrested four individuals for hacking telcos and carrying out sim swapping attacks. The group targeted cryptocurrency owners and are accused of stealing millions of US dollars. The four were arrested with the help of US authorities. They face prison sentences of up to 25 years each. Police in Montenegro have detained an individual believed to be an Iranian APT Group operator. The Iranian national, Amir Bharati was arrested last week in the Adriatic resort town of Kotor. He's believed to be a member of Silent Librarian, an APT that specialises in hacking and stealing private research from universities globally. He was arrested at the request of the FBI, which is now seeking his extradition to the US. The US State Department is offering a $10 million reward for information on the Russian hackers behind a recent wave of Signal and WhatsApp phishing campaigns. The US has linked the attacks to two hacking groups, tractors UNC4221 and UNC5792. Officials claim the former is linked to Russian military services, while the latter is associated with the FSB's Border Guards Division. Hackers are exploiting a vulnerability found in two software packages used to manage factories and production lines. The attacks against the PTC, Windchill and FlexPLM software began on June 18 and continued throughout last week. Attackers are exploiting an input validation bug in the software's web interface to deploy web shells to unpatched systems. The bug was added to Syskev database last week. A newly discovered indirect prompt injection attack can trick AI coding agents into infecting their owners with malware. The technique was discovered by Mozilla's AI ODIN Red Team. It relies on hiding malicious prompts inside DNS text records for domains that are listed in GitHub repository setup instructions. When an AI coding agent looks up the domain, the prompt is returned and injected. The technique bypasses security tools that scanned scan for payloads inside the repo itself. Around half of all new vulnerabilities are being enriched by NIST for the NVD portal under the agency's new enrichment rules. More than 6,700 vulnerabilities have been enriched for the NVD portal since mid April. NIST gave up on enriching all vulnerabilities after a large number of new vulnerabilities caused a backlog. The agency now prioritises bugs that are being exploited in the wild or in software used by the US government. And finally, Microsoft's security team has removed 119 malicious edge extensions from the Edge Add on store. The malicious extensions sought to steal user credentials, backdoor browsers and engage in advertising and search affiliate fraud. They were published through more than 90 different developer accounts, but shared infrastructure and parts of their code base. They all relied heavily on steganography to hide malicious commands and code. The extensions delivered genuine functionality but acted activated malicious modules days after they were installed. The Stego ad operation also ran malicious Chrome and Firefox extensions. More than 2.6 million users are believed to have downloaded the extensions and that is all for this podcast edition. Today's show was brought to you by our sponsor, Corelight. Find them@callight.com thanks for your company, Sam.