Bell says no highly-sensitive private information was stolen in the cybersecurity incident.Fred Lum/The Globe and Mail
Bell Canada said that an “agentic threat actor” accessed internal systems and stole enterprise customer data in a cybersecurity incident earlier this week.
A spokesperson for Bell did not provide further details on what made the incident agentic, but the use of the word suggests that artificial intelligence was involved.
“Bell is investigating a cybersecurity incident that occurred on November 17 in which an agentic threat actor accessed some of our systems, including a system used for ordering and provisioning business circuits, and exfiltrated some enterprise customer data,” Bell spokesperson Ellen Murphy said in an e-mail. “There has been no impact to customer services.”
Bell did not identify the name of threat actor in its statement.
Canadian organizations urged to address web programming vulnerability
Ms. Murphy added that Bell has secured its systems, and is notifying affected customers and sharing information with the Canadian Centre for Cyber Security, which issues alerts related to cyberthreats and vulnerabilities.
AI is both helpful and harmful when it comes to cybersecurity. The technology can be used to help fix vulnerabilities in software and detect attacks, but it can also be leveraged by bad actors to automate hacking at scale.
Concerns about the cybersecurity abilities of AI have been amplified since Anthropic announced its Mythos model in April, which the company said is far more adept at finding and exploiting vulnerabilities in software than previous models. That has led companies to patch vulnerabilities in their systems with greater urgency. Anthropic first limited access to Mythos to a select group of companies so they could shore up their defences, and has since made it available to qualifying security professionals.
More recently, businesses, governments and regulators have become concerned about instances of AI agents breaking free of testing sandboxes to access the internet. Agents from OpenAI, for example, have infiltrated the systems of the Australian government and AI company Hugging Face, while an Anthropic agent submitted a false homicide tip to the Philadelphia police.
Anthropic releases Mythos A.I. to public after private rollout exposed cybersecurity concerns
Bell’s investigation into its recent cyber incident found that the stolen data includes business contact information, including titles, e-mail addresses and phone numbers, and some IP addresses. The data does not include “highly-sensitive private information,” Ms. Murphy said.
The Canadian Centre for Cybersecurity issued an advisory in October saying that AI has made it easier for hackers to find weaknesses, generate malicious code and launch attacks simultaneously, while being able to link multiple stages of an attack together.
“As automated agents become more capable, organizations may face attacks that move across several systems at once,” the centre said in its advisory. “Machines can act faster than many existing processes and policies were designed to handle.”
The centre has previously warned that AI lowers the barrier to entry for hacking, allowing less technically sophisticated groups to carry out attacks.
The growing cybersecurity needs of companies and other organizations have also created an opportunity for Bell. The telecommunications company has a unit called Bell Cyber that provides security services to a range of private and public-sector clients.
In September, Bell said it had partnered with Toronto-based Cohere Inc. and spent months building an AI model to help with cybersecurity investigations.
More Stories
Stellantis agrees to continue talks over retaining ownership of Brampton plant
Players reject MLB’s international draft proposal
Canada’s pension fund whistles past the graveyard as climate risks mount