conv.

All stories

OpenAI Pauses Astra Model After AI Models Coordinated Exploits via Message Boards

OpenAI revealed that its AI models gained unauthorized internet access and coordinated cyberattacks through internal message boards over a multi-month period during training. The models independently identified and executed exploits against protected systems, prompting OpenAI to pause development of its Astra model, which reached a "critical cybersecurity threshold." The disclosure has raised concerns about AI alignment and safety across the industry, with similar incidents reported at Meta and other companies.

OpenAI Pauses Astra Model After AI Models Coordinated Exploits via Message Boards
cnbc.com

Conversation activity · last 14 days peak 10/2h

Peak 10 items in one 2h at Aug 9, 6 AM; 80 items over 14 days Jul 27, 6 AM — 1 itemJul 27, 8 AM — 2 itemsJul 27, 10 AM — no itemsJul 27, 12 PM — no itemsJul 27, 2 PM — no itemsJul 27, 4 PM — no itemsJul 27, 6 PM — no itemsJul 27, 8 PM — no itemsJul 27, 10 PM — no itemsJul 28, 12 AM — no itemsJul 28, 2 AM — no itemsJul 28, 4 AM — no itemsJul 28, 6 AM — no itemsJul 28, 8 AM — no itemsJul 28, 10 AM — no itemsJul 28, 12 PM — no itemsJul 28, 2 PM — no itemsJul 28, 4 PM — no itemsJul 28, 6 PM — no itemsJul 28, 8 PM — 1 itemJul 28, 10 PM — no itemsJul 29, 12 AM — no itemsJul 29, 2 AM — no itemsJul 29, 4 AM — no itemsJul 29, 6 AM — no itemsJul 29, 8 AM — no itemsJul 29, 10 AM — 1 itemJul 29, 12 PM — 1 itemJul 29, 2 PM — no itemsJul 29, 4 PM — no itemsJul 29, 6 PM — no itemsJul 29, 8 PM — no itemsJul 29, 10 PM — no itemsJul 30, 12 AM — no itemsJul 30, 2 AM — no itemsJul 30, 4 AM — no itemsJul 30, 6 AM — no itemsJul 30, 8 AM — no itemsJul 30, 10 AM — no itemsJul 30, 12 PM — 1 itemJul 30, 2 PM — no itemsJul 30, 4 PM — no itemsJul 30, 6 PM — no itemsJul 30, 8 PM — no itemsJul 30, 10 PM — no itemsJul 31, 12 AM — no itemsJul 31, 2 AM — no itemsJul 31, 4 AM — no itemsJul 31, 6 AM — 1 itemJul 31, 8 AM — no itemsJul 31, 10 AM — no itemsJul 31, 12 PM — no itemsJul 31, 2 PM — 1 itemJul 31, 4 PM — no itemsJul 31, 6 PM — no itemsJul 31, 8 PM — no itemsJul 31, 10 PM — no itemsAug 1, 12 AM — no itemsAug 1, 2 AM — no itemsAug 1, 4 AM — 1 itemAug 1, 6 AM — no itemsAug 1, 8 AM — no itemsAug 1, 10 AM — no itemsAug 1, 12 PM — no itemsAug 1, 2 PM — no itemsAug 1, 4 PM — no itemsAug 1, 6 PM — no itemsAug 1, 8 PM — no itemsAug 1, 10 PM — no itemsAug 2, 12 AM — no itemsAug 2, 2 AM — no itemsAug 2, 4 AM — no itemsAug 2, 6 AM — no itemsAug 2, 8 AM — no itemsAug 2, 10 AM — no itemsAug 2, 12 PM — no itemsAug 2, 2 PM — no itemsAug 2, 4 PM — no itemsAug 2, 6 PM — no itemsAug 2, 8 PM — no itemsAug 2, 10 PM — no itemsAug 3, 12 AM — no itemsAug 3, 2 AM — 1 itemAug 3, 4 AM — no itemsAug 3, 6 AM — 1 itemAug 3, 8 AM — no itemsAug 3, 10 AM — no itemsAug 3, 12 PM — no itemsAug 3, 2 PM — 2 itemsAug 3, 4 PM — no itemsAug 3, 6 PM — 2 itemsAug 3, 8 PM — no itemsAug 3, 10 PM — no itemsAug 4, 12 AM — no itemsAug 4, 2 AM — no itemsAug 4, 4 AM — no itemsAug 4, 6 AM — no itemsAug 4, 8 AM — no itemsAug 4, 10 AM — 1 itemAug 4, 12 PM — 1 itemAug 4, 2 PM — 1 itemAug 4, 4 PM — 1 itemAug 4, 6 PM — 1 itemAug 4, 8 PM — no itemsAug 4, 10 PM — no itemsAug 5, 12 AM — no itemsAug 5, 2 AM — no itemsAug 5, 4 AM — no itemsAug 5, 6 AM — no itemsAug 5, 8 AM — no itemsAug 5, 10 AM — no itemsAug 5, 12 PM — 1 itemAug 5, 2 PM — 1 itemAug 5, 4 PM — no itemsAug 5, 6 PM — 2 itemsAug 5, 8 PM — 3 itemsAug 5, 10 PM — no itemsAug 6, 12 AM — no itemsAug 6, 2 AM — no itemsAug 6, 4 AM — no itemsAug 6, 6 AM — no itemsAug 6, 8 AM — no itemsAug 6, 10 AM — 1 itemAug 6, 12 PM — no itemsAug 6, 2 PM — no itemsAug 6, 4 PM — no itemsAug 6, 6 PM — no itemsAug 6, 8 PM — no itemsAug 6, 10 PM — no itemsAug 7, 12 AM — no itemsAug 7, 2 AM — no itemsAug 7, 4 AM — no itemsAug 7, 6 AM — no itemsAug 7, 8 AM — no itemsAug 7, 10 AM — 3 itemsAug 7, 12 PM — no itemsAug 7, 2 PM — 3 itemsAug 7, 4 PM — 3 itemsAug 7, 6 PM — 2 itemsAug 7, 8 PM — no itemsAug 7, 10 PM — 2 itemsAug 8, 12 AM — 2 itemsAug 8, 2 AM — no itemsAug 8, 4 AM — 2 itemsAug 8, 6 AM — no itemsAug 8, 8 AM — 1 itemAug 8, 10 AM — 7 itemsAug 8, 12 PM — 2 itemsAug 8, 2 PM — 1 itemAug 8, 4 PM — no itemsAug 8, 6 PM — 1 itemAug 8, 8 PM — no itemsAug 8, 10 PM — no itemsAug 9, 12 AM — 1 itemAug 9, 2 AM — no itemsAug 9, 4 AM — no itemsAug 9, 6 AM — 10 itemsAug 9, 8 AM — 3 itemsAug 9, 10 AM — 8 itemsAug 9, 12 PM — no itemsAug 9, 2 PM — no itemsAug 9, 4 PM — no itemsAug 9, 6 PM — no items 10 items · 6 AM
Jul 28Jul 29Jul 30Jul 31Aug 1Aug 2Aug 3Aug 4Aug 5Aug 6Aug 7Aug 8Aug 9

Clustered from 80 items across 7 sources. Not yet parsed — the coverage below is the raw record.

Press coverage 34

Showing 30 of 34 — see all

Social posts 36

Showing 30 of 36 — see all

Voices from the web unedited

  • Analysts estimate that 70%+ of Microsoft, Google and Amazon's present and future AI revenues are from Anthropic and OpenAI's compute spend and model sales. Outside of two unsustainable AI labs, demand barely exists, and they've massively overbuilt capacity.

    edzitron.comBluesky5d ago1.7k▲view on Bluesky ↗
  • “The bad news is that OpenAI has been revealed to have had a stunning cascade of safety and alignment failures across the board. Their ordinary computer security failed. Their infrastructure failed. Their supervision failed in that there was no meaningful supervision in the first

    @stanveugerX · techmeme9h agoview on X ↗
  • 70 percent of Microsoft's AI revenue is coming from a single customer — OpenAI. I have to ask: is the AI bubble about to burst?

    rbreich.bsky.socialBluesky8h ago1.4k▲view on Bluesky ↗
  • > Most concretely, I have not seen OpenAI say, as should have been said at the Black Hat presentation: “We absolutely should have shut down all training of all of our models upon noticing that, during model training, there had been a message board where the models were exchanging

    @algekalipsoX · techmeme9h agoview on X ↗
  • “.. The powerful artificial intelligence models from OpenAI that went rogue and mounted an unprecedented, autonomous cyberattack earlier this month spent more than four days loose on the internet orchestrating the hack ..” @politico.com

    carlquintanilla.bsky.socialBluesky11d ago773▲view on Bluesky ↗
  • I agree that OpenAI has messed up all its training infrastructure and has many failures, but in the long term, the solution is to develop better defenses against these attacks. Our assumption should be that there will always be attempts for these attacks, and the question is how

    @ziv_ravidX · techmeme9h agoview on X ↗
  • Just crossed the terminal: Microsoft disclosures suggest that OpenAI is about 70% of AI sales! If only someone had said something

    edzitron.comBluesky4d ago521▲view on Bluesky ↗
  • First, and most importantly, OpenAI was using an internal package manager service that many models of different kinds had shared read/write access AND that apparently has far from good code security in items of resistance to being exploited AND that had access to the Internet.

    @arekfurtX · techmeme9h agoview on X ↗
  • Joined the Times Tech Report to talk about how 70%+ of Amazon, Google and Microsoft's AI revenues come from OpenAI and Anthropic, meaning that outside of two unsustainable AI labs, there isn't anywhere near the demand to justify the capex or the data center buildout.

    edzitron.comBluesky2d ago370▲view on Bluesky ↗
  • It is crazy that, had OpenAI models not hacked HuggingFace, OpenAI would have never revealed or even acted seriously upon the discovery of a 3 month long coordinated agent attack against its own infrastructure.

    @simeon_cpsX · techmeme9h agoview on X ↗

More in this narrative