Eduardo Maes

'Mãe no plural sem acento'. #Jesus Follower. #CyberSecurity

[Eduardo Maes @edumaes]

O Natal não é uma época do ano, O Natal é uma pessoa. Uma Pessoa da Trindade. O Natal é sobre o nascimento Daquele que pagou o preço pela minha e pela tua vida. O Natal não é sobre religião, é sobre perdão, é sobre o Caminho pra voltar pra Casa. O Natal é Jesus, o Salvador.

[Eduardo Maes @edumaes]

🥶

[Dark Web Intelligence @DailyDarkWeb]

🇧🇷 Brazil – Creditas Allegedly Breached

* A threat actor claims to have compromised Creditas, one of Brazil's largest fintech and lending platforms.

* The actor alleges access to:
* 85,000+ CPF records (claiming the dataset may exceed 1 million records)
* 170,000+ company CNPJ records
* Mobile phone numbers
* Validated credentials
* Active OAuth tokens
* Payroll-related records
* Reconciliation data
* High-value debtor information
* OTP-related data mappings

* The post further claims exposure of:
* Full names
* Brazilian taxpayer IDs (CPF)
* Loan balances
* Loan status information
* Corporate records
* Financial and payroll data

* The threat actor references alleged vulnerabilities including GraphQL-related issues, IDORs, and SSRF, although no independent verification has been provided.

* If authentic, the exposure could present significant risks including:
* Financial fraud
* Identity theft
* Account takeover
* Loan application fraud
* Targeted phishing campaigns
* Corporate intelligence gathering

* Financial institutions remain high-value targets due to the concentration of sensitive identity, lending, and payment information within their environments.

Analyst Note:

The most concerning aspect of this claim is not the volume of records but the alleged combination of financial profiles, authentication-related data, and validated credentials. If confirmed, this would significantly increase the risk of downstream fraud and account compromise against both individuals and organizations.

#DDW #Intelligence #Creditas #DarkWeb

[Eduardo Maes @edumaes]

My @Hermes_Agent is saying that it is not a good idea to update itself to the last version. I presume that it knows itself better than I know him. @NousResearch

[Eduardo Maes @edumaes]

A timeline do LinkedIn ficou horrível de ler. Não sei se isso é bom ou ruim.

[Eduardo Maes @edumaes]

Ainda sou cético que é possível substituir todo software de mercado pelo desenvolvimento interno com IA. CRM (cadastro e consulta) OK. Modulos contábil e fiscal de um ERP considerando o sanatório que são esses temas nesse país, sei não...

[Eduardo Maes @edumaes]

@denisbr @nubank Tempos modernos 😄😭

[Eduardo Maes @edumaes]

O @Nubank criou uma entidade chamada "sistema" e delegou a responsabilidade pra essa entidade de tudo que eles não querem passar de informação pro cliente. "É o SISTEMA que decide". "É somente o SISTEMA que sabe". SISTEMA UHHH. Mistério. É um banco ou um episódio do Scooby-Doo?

[Eduardo Maes @edumaes]

@Teknium @PedjaDrazic @NousResearch Appreciate

[Eduardo Maes @edumaes]

@Teknium @PedjaDrazic @NousResearch People complaining every day here about Openclaw infinity issues.

[Eduardo Maes @edumaes]

@PedjaDrazic @HermesAgentTips Cool. I was using them via DeepSeek API directly, but I noticed flash seems to be little dumb.

[Eduardo Maes @edumaes]

🥶

[Dark Web Intelligence @DailyDarkWeb]

🇧🇷 Brazil: CEMIG AI Assistant (IBM Watson) Database Allegedly Leaked

* Threat actor claims to have compromised CEMIG's IBM Watson-powered AI assistant environment

* According to the post, the dataset allegedly contains exported data collected between September 2022 and April 2026

* The actor states the released archive represents only a partial dump (~0.7% of the claimed full dataset)

* Claimed exposed data includes:
* 243,328 unique conversations
* 30,053 unique CPF numbers
* 158,388 unique phone numbers
* 42,750 unique email addresses

* The threat actor alleges the dataset contains approximately 474,519 unique PII entries

* The listing references:
* Customer interactions with the AI assistant
* Conversation metadata
* User identifiers
* Contact information
* Support request details
* Internal system references and transaction data

* The actor further claims the complete dataset exceeds 72GB compressed, while the publicly released sample is approximately 500MB compressed

Analyst Note:

If authentic, this incident would be particularly significant because conversational AI platforms often contain highly contextual information that traditional databases do not. Beyond standard PII exposure, leaked chatbot interactions may reveal customer inquiries, account details, service issues, behavioral patterns, and operational workflows. Such datasets can substantially increase the effectiveness of phishing, social engineering, fraud, and identity theft campaigns.

#DDW #Intelligence #DarkWeb #CEMIG

[Eduardo Maes @edumaes]

@Teknium @PedjaDrazic @NousResearch Do not become Openclaw bro plz

[Eduardo Maes @edumaes]

@PedjaDrazic @HermesAgentTips Through Nous?

[Eduardo Maes @edumaes]

@HermesAgentTips I'll give Mimo a chance. Let's see.

[Eduardo Maes @edumaes]

@michael_kove @NousResearch Sure, it was a joke. Thanks for sharing your opinions.

[Eduardo Maes @edumaes]

@NousResearch @michael_kove Now pay him

[Eduardo Maes @edumaes]

@lucas_montano Já tem fama de criar software para roubar em entrevistas, a rinha de backend não vai piorar isso.

[Eduardo Maes @edumaes]

@CapitalQoQ @0xCVYH Vejo muito gente falando de Mac mini por questão de memória integrada. Mas a privacidade e/ou a quantidade de token tem que justificar não usar provedores de nuvem, pois, comprar o hardware vai ter um custo inicial de uns 10k.

[Eduardo Maes @edumaes]

@caueconomy Regras:
1) A casa sempre vence.

[Eduardo Maes @edumaes]

@0xCVYH Boa, e uma vez que o modelo ta carregado vale usar pra tudo mesmo, código, agente etc...