WARPTECHNEWS · LAB

Home AI Business Tech Archive

WARPTECH LAB NEWS

Warptech Lab News aggrega le notizie più rilevanti da oltre 700 fonti internazionali, con classificazione AI, TL;DR sintetici e timeline cluster su singole storie.

Navigazione

Home
Archivio
Editor's Brief
Cerca
Il tuo account
Newsletter tech/AI

Informazioni legali

Privacy Policy
Termini di servizio
Cookie Policy

© 2026 Sparktech S.R.L. — Tutti i diritti riservati. Sito gestito e manutenuto da Sparktech S.R.L.

Sede legale: Corso Libertà 55, 13100 Vercelli (VC), Italia · P.IVA / C.F. 02835910023 · Contatti: admin@warptechlab.com

Storia in 13 fonti

OpenAI Details GPT-Red: An Internal Automated Red-Teaming Model That Beat Human Red-Teamers 84% To 13% On Prompt Injection

OpenAI details GPT-Red, an internal self-play red-teaming model that beat human red-teamers 84% to 13% and hardened GPT-5.6.

Raccontata da

technologyreview.com

the-decoder.com

cryptobriefing.com

siliconangle.com

newsbytesapp.com

marktechpost.com+1 altre

Confronto fonti

6 prospettive sulla stessa storia

AI · summaries

marktechpost.comStai leggendo5 h fa

OpenAI Details GPT-Red: An Internal Automated Red-Teaming Model That Beat Human Red-Teamers 84% To 13% On…

OpenAI details GPT-Red, an internal self-play red-teaming model that beat human red-teamers 84% to 13% and hardened GPT-5.6.

Timeline cronologica

mercoledì 15 luglio 2026·technologyreview.com
Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safer
Exclusive: The firm says it wants to future-proof its safety procedures and stay ahead of human attackers.
mercoledì 15 luglio 2026·decrypt.co
OpenAI Uses AI Red Team to Strengthen GPT-5.6 Against Prompt Injection Attacks - Decrypt
OpenAI said its automated red-teaming model, GPT-Red, uncovered vulnerabilities that were used to make GPT-5.6 more resistant to attacks.

siliconangle.com

1 g fa

OpenAI details GPT-Red, an AI that attacks its own models to find flaws - SiliconANGLE

OpenAI details GPT-Red, an AI that attacks its own models to find flaws - SiliconANGLE

Leggi questa versione → originale

the-decoder.com1 g fa

OpenAI is now using AI to attack its own AI, and it's working better than humans ever did

OpenAI's internal GPT-Red model finds successful attacks in 84 percent of test scenarios through self-play training. Human red teamers manage just 13 percent. The results feed directly into hardening models like GPT-5.6…

Leggi questa versione → originale

decrypt.co1 g fa

OpenAI Uses AI Red Team to Strengthen GPT-5.6 Against Prompt Injection Attacks - Decrypt

OpenAI said its automated red-teaming model, GPT-Red, uncovered vulnerabilities that were used to make GPT-5.6 more resistant to attacks.

Leggi questa versione → originale

theverge.com1 g fa

OpenAI trained an AI for red-teaming AI.

The new GPT-Red model “can break nearly all models it is pitted against,” according to an OpenAI blog post on Wednesday. OpenAI says it used GPT-Red to find vulnerabilities in GPT-5.6 Sol, a process that made it the…

Leggi questa versione → originale

cryptobriefing.com1 g fa

OpenAI strengthens GPT-5.6 against prompt injection attacks with internal AI red team

OpenAI's GPT-5.6 uses an internal AI red team called GPT-Red to achieve a 6x reduction in prompt injection failures, with a 0.05% failure rate against its

Leggi questa versione → originale

mercoledì 15 luglio 2026·

theverge.com

OpenAI trained an AI for red-teaming AI.

The new GPT-Red model “can break nearly all models it is pitted against,” according to an OpenAI blog post on Wednesday. OpenAI says it used GPT-Red to find vulnerabilities in…

mercoledì 15 luglio 2026·

the-decoder.com

OpenAI is now using AI to attack its own AI, and it's working better than humans ever did

OpenAI's internal GPT-Red model finds successful attacks in 84 percent of test scenarios through self-play training. Human red teamers manage just 13 percent. The results feed…

mercoledì 15 luglio 2026·

thenextweb.com

OpenAI built GPT-Red to hack its own AI, and hid it

OpenAI built GPT-Red, an in-house AI hacker that attacks its own models to harden GPT-5.6 against prompt injection, and it works too well to release.

giovedì 16 luglio 2026·

cryptobriefing.com

OpenAI strengthens GPT-5.6 against prompt injection attacks with internal AI red team

OpenAI's GPT-5.6 uses an internal AI red team called GPT-Red to achieve a 6x reduction in prompt injection failures, with a 0.05% failure rate against its

giovedì 16 luglio 2026·

siliconangle.com

OpenAI details GPT-Red, an AI that attacks its own models to find flaws - SiliconANGLE

OpenAI details GPT-Red, an AI that attacks its own models to find flaws - SiliconANGLE

giovedì 16 luglio 2026·

newsbytesapp.com

Meet GPT-Red, AI 'super-hacker' OpenAI uses to stress-test its models

OpenAI developed GPT-Red, a super-hacker LLM, to enhance its models' defenses against cyberattacks, contributing to the robustness of its latest release, GPT-5.6.

giovedì 16 luglio 2026·

ai4business.it

GPT-Red, come OpenAI rafforza la sicurezza degli agenti - AI4Business

OpenAI presenta GPT-Red: red teaming automatico, più difese contro prompt injection e nuovi effetti sul mercato dell’AI

giovedì 16 luglio 2026·

hwupgrade.it

GPT-Red, il super-hacker di OpenAI che buca gli altri modelli (e un distributore automatico)

L'azienda l'ha chiamato GPT-Red e lo usa come sparring partner per blindare i suoi modelli contro il prompt injection. Contro GPT-5.6 meno di un attacco su quattro va a segno, ma…

giovedì 16 luglio 2026·

t3n.de

Zu gefährlich für die Öffentlichkeit? Warum OpenAI GPT-Red unter Verschluss hält

Laut Angaben des Unternehmens ist der KI-Hacker zu leistungsfähig, um ihn öffentlich zur Verfügung zu stellen.

giovedì 16 luglio 2026·

marktechpost.com

OpenAI Details GPT-Red: An Internal Automated Red-Teaming Model That Beat Human Red-Teamers 84% To 13% On Prompt Injection

OpenAI details GPT-Red, an internal self-play red-teaming model that beat human red-teamers 84% to 13% and hardened GPT-5.6.

venerdì 17 luglio 2026·

aibusiness.com

OpenAI Unveils GPT-Red to Test AI Model Safety

While red teaming is standard practice, using humans and AI to test the security of new models is novel. Enterprises shouldn't let down their guard.