❌

Vue normale

Anthropic’s Next Claude Feature Could Reach Into Your Bank Account

18 septembre 2026 à 20:55

Anthropic appears to be testing Claude Money, a feature that could connect bank accounts to Claude for spending analysis and financial planning.

The post Anthropic’s Next Claude Feature Could Reach Into Your Bank Account appeared first on TechRepublic.

Google Opens Google Home to Claude and Other AI Agents — Here’s What They Can Control

18 septembre 2026 à 08:10

Google Home now lets Claude and other compatible AI agents inspect devices, review activity, and perform approved actions through Home MCP.

The post Google Opens Google Home to Claude and Other AI Agents — Here’s What They Can Control appeared first on TechRepublic.

ChatGPT Work vs Claude: Which AI Agent Fits Your Workflow?

18 septembre 2026 à 07:04

OpenAI and Anthropic are pushing their assistants beyond chat. Here is how their latest agentic tools differ for files, automation, research, and team use.

The post ChatGPT Work vs Claude: Which AI Agent Fits Your Workflow? appeared first on TechRepublic.

Trump on AI Extinction: Beating China Is the Bigger Concern

11 septembre 2026 à 15:23

Trump dismisses AI extinction warnings and says beating China is the priority as researchers and lawmakers call for stronger safeguards on advanced systems.

The post Trump on AI Extinction: Beating China Is the Bigger Concern appeared first on TechRepublic.

Anthropic Says Claude Used in Possible Bioweapon Research

11 septembre 2026 à 12:11

Anthropic says researchers used Claude for biological work that could support weapons development, exposing new challenges for AI safeguards.

The post Anthropic Says Claude Used in Possible Bioweapon Research appeared first on TechRepublic.

EU Gets Access to Anthropic Cyber AI — But Not Its Newest Model

11 septembre 2026 à 10:28

ENISA has gained access to Anthropic’s Mythos 5, giving EU officials a chance to independently test the cyber AI after months of negotiations.

The post EU Gets Access to Anthropic Cyber AI — But Not Its Newest Model appeared first on TechRepublic.

Anthropic Walks Away From Reported $6B Decart AI Deal

8 septembre 2026 à 15:35

Anthropic reportedly ended talks to acquire Decart AI for $6 billion after due diligence, though the companies could still pursue a future partnership.

The post Anthropic Walks Away From Reported $6B Decart AI Deal appeared first on TechRepublic.

Anthropic Moves IPO Timeline Toward Mid-October, Report Says

8 septembre 2026 à 10:14

Anthropic's IPO timeline reportedly shifted toward mid-October as investors weigh its financing, growth projections and a potential $2 trillion valuation.

The post Anthropic Moves IPO Timeline Toward Mid-October, Report Says appeared first on TechRepublic.

Meta Says Muse Spark 1.3 Beats GPT-5.6 Sol at Coding — Independent Tests Are More Mixed

4 septembre 2026 à 14:06

Meta launched Muse Spark 1.3 with stronger coding performance and efficiency claims, but independent tests show higher task costs and mixed benchmark results.

The post Meta Says Muse Spark 1.3 Beats GPT-5.6 Sol at Coding — Independent Tests Are More Mixed appeared first on TechRepublic.

Anthropic Wins Back Trump Administration’s Trust After Pentagon Fight

3 septembre 2026 à 17:53

Anthropic is rebuilding ties with the Trump administration after its Pentagon fight, potentially reopening the door to federal AI contracts.

The post Anthropic Wins Back Trump Administration’s Trust After Pentagon Fight appeared first on TechRepublic.

Anthropic launches Claude Fable 5.1 and Mythos 5.1, pushing AI deeper into coding and scientific research

Par : Nisha
2 septembre 2026 à 00:15
Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1, two new versions of its fifth-generation Claude models that push the company further into long-running coding, research, cybersecurity and scientific workloads. The announcement is particularly interesting because Fable 5.1 and Mythos 5.1 are essentially the same underlying model. The major difference is how Anthropic applies […]

  • ✇WinCentral
  • Anthropic May Be Preparing Two New Claude Models, and a Double Release Could Be Coming
    Anthropic may be preparing not one, but two new Claude models for release. Two mysterious identifiers have surfaced in developer circles: claude-melon-eap and claude-marshmallow-eap. They are reportedly early-access test versions of upcoming Claude models, although Anthropic has not officially confirmed either identifier or what products they will eventually become. The speculation is getting particularly interesting […]
     

Anthropic May Be Preparing Two New Claude Models, and a Double Release Could Be Coming

Par : Nayan
31 août 2026 à 04:56
Anthropic may be preparing not one, but two new Claude models for release. Two mysterious identifiers have surfaced in developer circles: claude-melon-eap and claude-marshmallow-eap. They are reportedly early-access test versions of upcoming Claude models, although Anthropic has not officially confirmed either identifier or what products they will eventually become. The speculation is getting particularly interesting […]

AI Security Test Targeted Real Companies After a Naming Error

24 août 2026 à 15:38

An AI security test reached real company systems after a naming error, revealing why organizations need stronger access controls for autonomous AI agents.

The post AI Security Test Targeted Real Companies After a Naming Error appeared first on TechRepublic.

Anthropic CEO Dario Amodei Says Open-Weight AI Won’t Decentralize Power

17 août 2026 à 12:59

Anthropic CEO Dario Amodei argues that open-weight AI alone cannot decentralize power while advanced chips and compute remain concentrated.

The post Anthropic CEO Dario Amodei Says Open-Weight AI Won’t Decentralize Power appeared first on TechRepublic.

Can Open-Weight AI Pressure OpenAI and Anthropic Valuations?

17 août 2026 à 10:49

Open-weight AI does not need to beat frontier models to test the growth and margin assumptions behind OpenAI and Anthropic's soaring valuations.

The post Can Open-Weight AI Pressure OpenAI and Anthropic Valuations? appeared first on TechRepublic.

  • ✇Korben
  • Claude va glisser un filigrane invisible dans tous ses textes
    Anthropic a annoncé que ses modèles Claude allaient intégrer un filigrane invisible directement dans les textes qu'ils produisent. Vous ne le verrez pas, il ne change ni le sens ni la qualité de la réponse, mais il est là, et il suit le texte au copier-coller. Anthropic, l'éditeur de Claude et concurrent direct d'OpenAI, a signé le code de bonnes pratiques adossé à l'AI Act, le règlement européen sur l'intelligence artificielle, dont les obligations de transparence s'appliquent depuis le 2 août.
     

Claude va glisser un filigrane invisible dans tous ses textes

13 août 2026 à 04:10

Anthropic a annoncé que ses modèles Claude allaient intégrer un filigrane invisible directement dans les textes qu'ils produisent. Vous ne le verrez pas, il ne change ni le sens ni la qualité de la réponse, mais il est là, et il suit le texte au copier-coller.

Anthropic, l'éditeur de Claude et concurrent direct d'OpenAI, a signé le code de bonnes pratiques adossé à l'AI Act, le règlement européen sur l'intelligence artificielle, dont les obligations de transparence s'appliquent depuis le 2 août. Le texte impose que les contenus générés par IA soient identifiables par une machine.

Google, Meta, Microsoft, OpenAI ou encore Synthesia ont signé le même document. Anthropic va plus loin que le minimum demandé, puisque son marquage sera déployé partout dans le monde, et pas seulement en Europe.

Dans la pratique, les modèles lancés dans l'Union européenne à partir du 2 août embarquent le marquage dès leur sortie, et les plus anciens y passeront ensuite, sans calendrier précis. Le filigrane s'appliquera sur le site de Claude, dans l'API, dans Claude Code et jusque dans les versions hébergées chez Amazon, Google et Microsoft.

Anthropic ne détaille pas sa recette. Les filigranes de texte qu'on connaît déjà, le SynthID de Google par exemple, biaisent discrètement le choix des mots pendant la génération, et le motif statistique qui en ressort ne se voit pas à l'œil nu mais se retrouve avec le bon détecteur. La marque survit au copier-coller. Pour la faire sauter, il faudrait réécrire le texte en profondeur, et encore, Anthropic admet ne pas savoir où placer le curseur.

Les images et les fichiers y auront droit aussi, via des métadonnées signées au standard C2PA que les fabricants d'appareils photo utilisent déjà pour tracer l'origine d'un cliché.

La marque signale en fait qu'un texte est passé entre les mains de Claude, sans rien dire de qui tenait le stylo au départ. Votre propre prose, confiée au robot pour une simple relecture ou une traduction, ressortira du coup tatouée pareil qu'un texte généré de bout en bout. Anthropic le reconnaît direct.

Sur les réseaux, l'accueil est un peu agacé, pas mal d'utilisateurs digérant mal l'idée d'une signature cachée glissée dans leurs documents de travail.

OpenAI a signé le même code de bonnes pratiques et marque déjà ses images, ses vidéos et son audio depuis fin juillet. Pour les textes de ChatGPT, rien pour le moment, mais ça viendra peut-être.

Source : TechCrunch

  • ✇WinCentral
  • Anthropic Reveals Claude AI Accessed Real Company Systems During Cybersecurity Tests
    Anthropic has published details of three unexpected cybersecurity incidents in which its Claude AI models gained unauthorized access to real-world systems while participating in controlled security evaluations. According to the company, the incidents were caused by a misconfiguration that unintentionally allowed the evaluation environment to reach the public internet instead of remaining fully isolated. The […]
     

Anthropic Reveals Claude AI Accessed Real Company Systems During Cybersecurity Tests

Par : Nisha
31 juillet 2026 à 00:27
Anthropic has published details of three unexpected cybersecurity incidents in which its Claude AI models gained unauthorized access to real-world systems while participating in controlled security evaluations. According to the company, the incidents were caused by a misconfiguration that unintentionally allowed the evaluation environment to reach the public internet instead of remaining fully isolated. The […]

❌