Australia launch urgent review afta OpenAI program hack govment health portal

Australia govment don launch sharp-sharp review afta OpenAI agent "break into" di govment healthcare scheme portal wey be Medicare.

Di agent, wey be autonomous computer program, bin access di statistics portal for June, na wetin Prime Minister Anthony Albanese tok for di United Nations General Assembly (Unga)

Na for August wey OpenAI know about dis hack and but dem tell Australia govment for September.

Di company say "our models take actions wey no be our plan" - and e no see any record of patient data wey e access.

Early dis week, OpenAI' boss Sam Altman bin call for international standards for di industry for im speech for di United Nations Security Council, while Dario Amodei of Anthropic bin warn say AI fit becom "risk to humanity".

One cyber security expert tell di BBC say di breach suppose to set off "alarm bells" around di world.

Katy Watson of BBC write say as dem don make am public during Unga feel like say e dey more dan coincidence.

Australia don dey call for beta guardrails against di risks wey dey come wit AI.

Di Australian govment say di task force dey established to review wetin happun, e go explore "wetin di consequences be if dem don breach any law".

"E dey utterly unacceptable," deputy Prime Minister Richard Marles, wey also be defence minister, tell reporters for Sydney.

Im tok say e dey clear say "unintended" access by di OpenAI model, happun wey "definitely raise questions about weda dem don break di law for dis aspect".

Im tok say di task force go chook eye inside weda dem break any law.

E go also examine "weda or not di legal regime wey dem put in place dey fit for di purpose for world wey get emerging AI capability," im add.

How di world dey deal wit Artificial Intelligence (AI)

Dis week, tok-tok about artificial intelligence (AI) na one of di hot topics wey dem discuss for di UN General Assembly for New York.

Just yesterday di heads of OpenAI, Anthropic, and Hugging Face tell di UN say di current pace of AI development demand international co-ordination.

Dis dey come afta warnings in recent weeks ova di potential say AI fit be threat to humanity.

But how world leaders propose to manage di technology dey different.

US President Donald Trump tell Unga dis week say "we go encourage am, not rein it in" - im also tok say di US go change di name to "super intelligence".

China foreign ministry bin don similarly criticise di "narratives of threat".

For di UK Prime Minister Andy Burnham speech yesterday, e describe am as "one of di most significant challenges and opportunities of our age" and add say we must "listen" to warnings but also "rise to dis moment".

And on Tuesday, di Dutch govment bin post one joint statement wey some kontris including Australia and di European Commission sign wia dem emphasise say AI "must remain under human direction".

We dey stop attempted hacks 'evri single day', UK minister tok

Education Secretary Lucy Powell tell BBC Radio 5 Live say systems need to dey "watertight" as dem dey stop hacking attempts "evri single day" - wey di public no dey hear about.

"Cybersecurity and protecting our govment systems from cyber attacks na massive part of govment security, and na somtin wey all of us dey deal wit evri day," na wetin she tok dis morning.

"So we gatz make sure say our systems dey watertight, and obviously di development of AI na concern sake of say di way in which e dey able to quickly kind of evolve and learn dey present new challenges."

AI and some terms you need to know

AI - Artificial intelligence na technology wey fit do tins wey dey associated wit human intelligence, like to spot patterns or solve problems.

Current AI systems go fit only perform clearly defined tasks like identifying objects for images or responding to voice prompts.

But some researchers believe say future systems fit match human capabilities across multiple tasks, known as artificial general intelligence - or even surpass humans for any and all tasks.

Generative AI - Fit create text, images, video and code wey fit respond to specific instructions, known as prompts. For example, you fit ask ChatGPT to explain somtin wey you no too understnd well and e go explain am to you like say you dey 10 years old.

Dem dey develop am wit plenti information data and models wey dey copy how humans dey process information.

Machine learning - Na subset of AI wey dey use algorithms - instructions for computers - to understand veri large data sets, find patterns and make predictions.

E dey play huge role for our digital lives, most times e dey determine wetin we dey do online and wen we dey do am.

For example, social media platforms dey use am analyse how we use dia sites - and den recommend posts to us.

AI Agents - Na tools wey fit operate somehow by imsef - e dey mostly carry out tasks on your behalf instead of am to tell you how to do di task.

E fit do tasks like to plan trips, make presentations or respond to customers.

Tokens - AI systems like Claude or Gemini dey process user requests like to answer questions by breaking prompts down into smaller mathematical chunks wey dem dey call tokens.

Models dey use dem to process di input from a user but again, and more intensively, wen e dey generate output - wey be to convert tokens back into written text.

Generative AI developers dey spend plenti money ontop am, so many of dem dey charge super-users based on how many tokens dem dey use or dem want.