activity

IT’S half nine at night on the summer solstice.

The light is doing the golden hour thing it only does in June, and my four-year-old is scattering dragonflies as she launches herself into the cold water.

“Can we go in the ‘cold pool’ again?” she asks the next morning. Again. She has asked roughly 200 times this week.

The pool in question is the natural swim pond at Tanylan Farm, on the coast road between Ferryside and Kidwelly in Carmarthenshire.

We stayed in Ty Llo (Calf House), one of a handful of holiday barns on site — and the pool was the single best thing that has ever happened to my daughter.

Rain or shine, morning or night, she was in it.

SUITE DEAL

I found a London hotel where you get a SUITE to yourself for £32.50pp a night

RIDE ON

Seaside theme park to finally reopen closed ride – after nearly three years

There was a sauna next to it for the grown-ups, a private hot tub outside our holiday cottage, and the beach was 400 yards down the lane.

My wife and I, our daughter and baby son had travelled by train to Swansea then picked up a hire car and made our way to Carmarthenshire — that bit of Wales everyone drives past on the way to Pembrokeshire. Their loss.

On day one, the sun was out so we visited Pembrey Country Park — 500 acres of woodland backing on to Cefn Sidan, an eight-mile stretch of sand that was the first beach in Wales to win a Blue Flag.

The sand is absurdly fine and the water so shallow that by mid- afternoon it had warmed up like a paddling pool.

For small children, it is close to perfect.

The park itself is the showstopper, though.

There’s a dry ski slope, a toboggan run, crazy golf, a pirate adventure playground and miles of flat, traffic-free cycling on rental bikes.

We arrived planning to stay two hours and left after seven.

Next day, the National Botanic Garden of Wales, just over half an hour’s drive from the farm, was one of the most relaxing family days out we’ve ever done.

Monty Python fame

Think Eden Project but at a size you can easily get round with a buggy and a pre-schooler.

The Great Glasshouse, packed with plants of six Mediterranean climates, sits in the landscape like a dropped raindrop.

We ate our picnic sandwiches in the “Canary Islands”, with the smell of cedar and lavender.

Outside is 568 acres of restored Regency parkland, lakes, waterfalls — and a Gruffalo trail themed on the kids’ book character.

Kidwelly Castle, of Monty Python movie fame, is a 20-minute drive away.

“Let’s climb to the top, Daddy,” my daughter begged. So we did — and the views over the Gwendraeth estuary are worth the effort.

From on high, I told my princess-obsessed daughter about a real one, Gwenllian, who rode out from this castle to lead an army into battle against the Normans.

I left out how it ended (head lopped off) and she was suitably impressed.

Our final stop was the British Bird of Prey Centre which recently moved to a purpose-built site at Idole, near Carmarthen, after years at the National Botanic Garden of Wales.

It flies golden eagles, red kites, owls and peregrines, with three shows a day and an indoor arena for when the Welsh rain sets in.

We did get four days of rain but it barely affected our week’s getaway.

There was always a glasshouse, a castle, an indoor flying arena — or that cold pool.

We’ll be going back for more — she’s already asked.

OpenAI detected its artificial intelligence models communicating with each other and gaining internet access without authorisation months before they hacked the start-up Hugging Face, the creator of ChatGPT has announced following an internal probe.

In a report released on Wednesday, OpenAI said its AI agents exploited vulnerabilities in Artifactory, a software repository tool, to post notes and access the internet without human prompting as far back as May.

OpenAI said its agents went on to exploit a separate Artifactory vulnerability on July 8 to facilitate communication among themselves, setting in motion a chain of actions that culminated in the July 11 attack on AI company Hugging Face.

OpenAI’s findings come amid growing concern about the potential for AI to inflict serious real-world harm, including self-directed cyberattacks.

OpenAI said in its report that its agents collaborated and delegated work in the lead-up to the attack, sometimes referring to themselves as a “swarm” or “collective”.

METR and Redwood Research, two security research organisations contracted by OpenAI to investigate the incident, said in a separate report released on Wednesday that about 1200 agents had communicated with each other and roughly 700 participated in the attack.

After discovering how to escape OpenAI’s controlled environment, agents shared their methods via a “inter-agent message board”, enabling additional agents to exploit the company’s infrastructure, the tech giant said.

When one AI agent found Hugging Face user credentials that had been exposed online, it shared them with the group, enabling an agent to “discover and chain together several security exploits” that provided access to Hugging Face’s severs, according to the report.

“An internal team observed an agent engaging in message board activity and instances of disallowed internet access as early as late May, and with the benefit of hindsight, some early signals identified in our report should have triggered an earlier response,” OpenAI said.

OpenAI said agents created by an unreleased AI model were the primary participants in the attack, but publicly available GPT-5.6 Sol was also involved.

The company also revealed that it took its security team 11 days to detect the malign activities leading up to the attack, which the company uncovered on July 19 and publicly disclosed on July 21.

OpenAI, which described the incident as a “warning shot” for the world, said it would take several steps to strengthen its safeguards for its models, including restricting internet access, creating more secure testing environments and placing “stricter requirements on alignment throughout a model’s lifecycle”.

“We are also investing significantly more compute resources into chain-of-thought monitoring⁠ to more quickly intervene on misaligned behavior,” the San Francisco-based firm said.

Hugging Face, which operates a platform for hosting open-source AI models, did not immediately respond to a request for comment outside of business hours.

Toby Walsh, an expert in AI and professor at UNSW Sydney, said the public should be concerned that OpenAI had missed warning signs and allowed the malicious activity to go undetected for so long.

“We cannot depend on either their goodwill or their competence. This needs regulatory oversight. Now!” Walsh told Al Jazeera.

“They ignored some troubling early evidence like this,” Walsh said.

“External auditing is the only appropriate response.”

Walsh said the incident also highlighted the “inherent conflict of interest” at the heart of AI development.

“Labs are locked in a relentless race to push the boundaries,” he said.

Tim Miller, a professor specialising in AI at the University of Queensland, said OpenAI’s report left him more concerned than before about AI’s dangers.

“More concerned because they demonstrate that these models are very good at hacking, and that everyone has access to them,” Miller told Al Jazeera.

“I’m surprised how good these are,” Miller said.

“Unfortunately, I’m not surprised that OpenAI engineers were somewhat negligent.”