Warning shot or publicity stunt - how worried should we be about the OpenAI hack?

Warning shot or publicity stunt - how worried should we be about the OpenAI hack? By Joe Tidy Cyber correspondent, BBC World Service Hugging Face announced on 16 July that it had been hacked by a cyber transgression wielding enormously almighty AI. The bombshell announcement was afloat of scary, highly method terms: "a swarm of sandboxes", "agentic attacker", and "self-migrating bid and control". Hugging Face said the hack was antithetic from thing it had handled earlier due to the fact that it was done astatine superhuman velocity by an AI with small oregon nary quality guidance. The AI performed 17,000 actions successful little than 2 days, successfully breaching the ample affluent tech institution to bargain secrets. It near the tech satellite successful shock. Hugging Face researchers guessed the mysterious attackers had utilized 1 of the large AI models but they had nary thought who oregon wherever the criminals were. The perplexed institution contacted the constabulary and investigations commenced. Commentators and analysts took to their podcasts and societal media accounts to conjecture which cyber transgression radical or federation authorities hacker mightiness beryllium down it. Then connected Wednesday, astir a week aft Hugging Face raised the alarm, the existent culprit was unmasked. It was ChatGPT. The Scooby-Doo-style uncover was made adjacent much bizarre and worrying due to the fact that OpenAI said its bot did the full happening connected its own, without permission. The steadfast said it each went down during a trial of its tech's hacking skills. Two caller versions of ChatGPT, designed to beryllium maestro hackers, broke retired of a supposedly unafraid trial situation and gained entree to the internet. They past attacked Hugging Face to get entree to the accusation to assistance them ace their exam. OpenAI issued a property merchandise explaining what had happened and said it was "partnering with Hugging Face" to code the information incidental and stock lessons learned. Since then, determination has been fierce statement astir the incident. Was it genuinely a stark informing astir the aboriginal of AI? Or was it a publicity stunt by OpenAI to amusement disconnected however almighty their models are? It's the benignant of scare selling AI companies person been accused of for years and, since the overmuch discussed motorboat of Anthropic's Mythos model, cyber-security prowess has been a focal point. One of the apical comments connected OpenAI brag Sam Altman's X station astir the incidental summarises this scepticism: "If y'all can't recognize that this was written to purely brag astir the exemplary past I don't cognize what to archer you." Cyber-security advisor Daniel Card said sarcastically on LinkedIn: "Isn't it fortunate [that] retired of the millions of sites that got pwn3d [hacked], OpenAI managed to pwn idiosyncratic who besides could payment from the selling exposure…" For some, the communicative is much conspiracy play than sci-fi thriller. The connection is: "Aren't my AI tools truly powerful? Buy them truthful you tin support yourself from different people's AI attacks." We can't cognize the truth, but the opposing constituent of presumption posed by different commentators is conscionable arsenic dramatic. Is this a motion that OpenAI made a perchance unsafe mistake successful judgement and planning? I've covered tons of AI stories, including