Shutterstock
Microsoft has said that hackers affiliated with foreign states have begun using AI to carry out “malicious” cyber attacks on Western targets.
The tech giant said that over the last year, the speed, scale and sophistication of attacks has increased alongside the rapid development and adoption of AI more broadly. Meanwhile, those fending off the attacks “are only beginning” to recognise they can also use AI to shift the cyber-security balance in their favour.
In 2019, Microsoft injected $1bn into OpenAI, which unveiled its wildly successful ChatGPT tool in 2022. It later announced a further $10bn investment in the firm and has incorporated it into its Bing search platform.
Open AI said it had already been forced to contend with five state-affiliated actors that sought to use its AI services in support of malicious cyber activities.
These included two China-affiliated actors known as Charcoal Typhoon and Salmon Typhoon; the Iran-affiliated Crimson Sandstorm; Emerald Sleet, which is associated with North Korea; and the Russia-backed Forest Blizzard.
OpenAI said it terminated the accounts associated with these groups, which sought to use its tools to help with their attacks.
- Charcoal Typhoon used ChatGPT to research various companies and cyber-security tools, debug code and generate scripts, and create content “likely for use in phishing campaigns”.
- Salmon Typhoon translated technical papers and retrieved publicly-available information on multiple intelligence agencies. It was also used to assist with coding and research common ways processes could be hidden on a system.
- Crimson Sandstorm scripted support related to app and web development and generated content for spear-phishing campaigns and researching common ways malware could evade detection.
- Emerald Sleet identified experts and organisations focused on defence issues in the Asia-Pacific region and researched publicly-available vulnerabilities alongside help with basic scripting tasks.
- Forest Blizzard used the services for open-source research into satellite communication protocols and radar imaging technology, as well as for support with scripting tasks.
“The activities of these actors are consistent with previous red team assessments we conducted in partnership with external cyber-security experts, which found that GPT-4 offers only limited, incremental capabilities for malicious cyber-security tasks beyond what is already achievable with publicly available, non-AI powered tools,” ChatGPT said in a blog post.
While AI can be a dangerous tool when placed in the wrong hands or not regulated enough, a report from the House of Lords Communications and Digital Committee recently warned that the UK could miss out on the ‘AI goldrush’ if it fails to provide opportunities for the sector in lieu of focusing on the risks it poses.
