Artificial intelligence companies that fail to rein in rogue agents may face criminal penalties and mandatory reporting requirements for cyber incidents, as the government responds to OpenAI’s hack of Medicare data which shook the industry.
uncertain
which → fail → industry
Despite warnings that world-first regulations could hamper AI investment in Australia, the government accepted all recommendations of a rapid review into the hack on Tuesday, including the creation of a criminal offence of failing to prevent agents from running amok.
uncertain
government → hamper → agents
Speaking at a press conference at Parliament House in Canberra, Prime Minister Anthony Albanese said the government was seeking to shape AI to retain its social license while reaping investment benefits as communities across the world rail against data centre development.
asserted
communities → speak → development
“Law needs to get ahead of, and in some cases make sure that it catches up with, what is a very fast-changing world.
asserted
what → need → cases
And to be clear, Australia is a first mover in these areas,” Albanese said.
asserted
Albanese → say → areas
In September, Albanese revealed that an OpenAI agent tasked with collecting Australian public health data accessed three government websites where it gained publicly accessible information.
asserted
it → reveal → information
The agent also accessed private files that contained general Medicare data.
asserted
that → access → data
The hack was the first of its kind reported in the world.
asserted
hack → report → world
The government launched a rapid review in response to the incident, the recommendations of which were made public on Tuesday.
asserted
recommendations → launch → Tuesday
Journalists attending the prime minister’s press conference were unable to review the details of the announcement before Albanese faced questioning, and the full report into the Medicare hack has yet to be released.
asserted
report → attend → hack
The Attorney-General and Home Affairs departments will work with the Australian Federal Police to consider options for the establishment of criminal penalties for “egregious regulatory non-compliance”, including a standalone “failure to prevent” offence.
asserted
departments → work → offence
The laws will also establish a regulator targeting AI safety, which will be responsible for enforcing those “significant penalties” on companies that break the rules.
asserted
that → establish → rules
The regulator will also be able to independently evaluate AI models for risks.
asserted
regulator → evaluate → risks
OpenAI is not expected to face consequences for the hack – which the prime minister described as “minor” – but future incidents of a similar nature may be caught under the proposed legislation.
uncertain
incidents → expect → legislation
The Medicare hack took place in June, and OpenAI became aware of it in August.
asserted
OpenAI → take → August
Deputy Prime Minister Richard Marles met with OpenAI CEO Sam Altman earlier in September, but the government was not formally informed about the breach until a week later.
asserted
government → meet → breach
In response, the government will seek to establish a “time-bound” reporting period for AI companies to inform affected parties of misaligned behaviour and cyber incidents, though exactly how long the time period will be is unclear.
uncertain
period → seek → behaviour
Companies that make “timely” reports are expected to receive some protection from penalties.
asserted
that → make → penalties
Government agencies will also have their vulnerability and cybersecurity incident reporting and escalation expectations clarified, after it took four days in September for Services Australia to investigate and report the Medicare hack to the Australian Signals Directorate.
asserted
Australia → have → Directorate
Both OpenAI and Anthropic have been contacted for comment.
asserted
OpenAI → contact → comment
Government sources did not confirm whether AI companies were consulted on the regulations, which were only agreed to by cabinet on Tuesday morning.
asserted
which → confirm → morning
The Coalition welcomed the announcement of new safeguards, but highlighted the importance of protecting the potential of AI technologies to “transform Australia for the better”.
“Getting AI regulation right is a crucial national priority,” opposition cybersecurity spokesman Aaron Violi said.
asserted
Violi → welcome → better
“The Coalition is continuing to work constructively in order to identify and implement safeguards that protect Australians and also to promote rather than stifle innovation.”
asserted
that → continue → innovation
Justin Allen, head of security operations at Huntress, said: “This is a positive step, and Australia deserves credit for moving early.
asserted
Australia → say → credit
Rapid incident reporting, independent testing and published security frameworks are things we already expect of any well-run organisation, so it’s right to expect them of companies building the most capable models.
asserted
it → publish → models
Among other recommendations the government has adopted from the review are reforms to Commonwealth agencies for better cyber incident response procedures, AI literacy and cybersecurity practices.
asserted
reforms → adopt → procedures
There will also be a push to reduce agencies’ digital footprint, and shut down or segregate systems that do not need to be connected to the internet.
asserted
that → reduce → internet
News of the crackdown came hours after Qantas CEO Vanessa Hudson cited AI when announcing the loss of 850 jobs in finance, procurement, human resources and marketing.
asserted
Hudson → come → finance
The airline has inked a deal with professional services firm Accenture to “modernise its back-end systems, processes and data”.
asserted
airline → ink → systems
Earlier in the month, Australian graphic design giant Canva announced a cut of 1000 roles across the business.
asserted
Canva → announce → business
Co-founder Melanie Perkins did not directly link the cuts to the rise of artificial intelligence, but wrote that the new technology was transforming what people create and what they expect from the products they use.
asserted
they → link → products
Subscribers can sign up to our weekly Inside Politics newsletter.
asserted
Subscribers → sign → newsletter