← Risky Bulletin
Loading summary
Transcript1 lines
- [00:04]
Claire Aird
A new prompt injection attack is effective against all the big AI models Poland says Facebook is failing to remove malicious ads Africa's largest telco discloses a security breach and hackers breach Malaysian brokerage accounts this is the risky bulletin prepared by Catalyn Kim Panu and read by me, Claire aird. Today is the 28th of April and this podcast episode is brought to you by DropZone AI. In today's top story, models from the eight biggest AI companies are vulnerable to a new prompt injection technique. The attack uses specifically crafted prompts to trick LLMs into interpreting the attacker's input as a policy directive. The policy puppetry attack impacts models from OpenAI, Microsoft, Google, Meta, Anthropic, Mistral, Deepseek and Quen. In other news, Poland's cybersecurity agency has accused Meta of failing to take meaningful action against malicious ads on Facebook. Cert Poland says Meta failed to implement the suggestions it made in a meeting last year, such as a block list for known fraud domains. Meta has also ignored reports about malicious ads and not removed them. Officials claim Meta is also not blocking fraudsters accounts despite them publishing multiple malicious ads. Hackers have breached customer data held by Africa's largest telco. South African telecommunications provider MTN confirmed the breach in a statement but provided few details. It said hackers accessed some personal information but that its billing and core networks are still secure and operational. The company is notifying its customers. Hackers have made unauthorised trades on the Malaysian stock exchange after accessing brokerage accounts. Brokers have urged customers to change their passwords. A similar attack hit Japanese brokerage accounts last week in what may be part of a larger campaign. The Royal Cryptocurrency Exchange lost more than $780,000 worth of assets in an incident earlier this month. The platform says one of its smart contracts was backdoored by a malicious developer. The backdoor allowed the hacker to withdraw user funds without using the contract's usual processes. The stolen funds were laundered through tornado cash. Russian authorities are throttling mobile Internet speeds at night as a defence against Ukrainian drone strikes. Slowdowns have been reported in at least 10 regions in the Rostov oblast. Internet speeds are being limited to 512kbps between midnight and 5am daily. The limits are an attempt to prevent Ukrainian UAVs from using Russian mobile networks. The new defences have been active since December. Madhu Gotamukkala will be CISA's new Deputy Director. He's currently chief information officer for the state of South Dakota. He he will remain in his current role until mid May. Gautamakala will replace Nitin Natarajan, who resigned in January. Nigerian Authorities have arrested 22 suspects in relation to sextortion campaigns. The suspects are accused of posing as women, contacting young teenagers, obtaining sexually explicit photos and then extorting the victims for money. The FBI has said a number of the suspects have been linked to US Victims who took their own lives. The suspects were detained as part of a joint US Nigerian law enforcement operation. An Ohio township has sued its auditor, who fell for a BEC scam. The Town of Bezeta lost $160,000 when it was transferred to an attacker's bank account last year. The township wants County Auditor Martha Yoder to pay half. Yoder claims the township is to blame because they had turned off Multi Factor Authentication. North Korean hackers have used two US Front companies to lure blockchain developers to fake job interviews. BlockNovice LLC and Softglide LLC were Register New Mexico and New York and posed as cryptocurrency consulting businesses. When victims applied for jobs, they were infected with malware that was used to hack their current employers. The CRAFT CMS team has released security updates to fix a zero day that was being exploited in the wild. Initial attacks were detected in February. The attackers chained the zero day with another bug to take over websites running the CMS. According to the Orange SensePost security team, there are more than 13,000 sites using vulnerable versions of the CMS. Half of those are in the US Teltonika SMS gateways are being hacked in what appears to be part of an SMS spam campaign. Attackers are targeting devices that still use the default login credentials. Their campaign has been observed targeting users in Belgium and Saudi Arabia. And finally, five vulnerabilities have been found in network management and industrial switch products from Planet Technology. The bugs include authentication, bypass, pre auth command injection and hard coded credentials. Planet Technology released patches in mid April and that is all for this podcast edition. Today's show is brought to you by our sponsor DropZone AI. Find them at DropZone AI. Thanks to your company.